.github/workflows/pylint.yml

@@ -1,23 +0,0 @@
-name: Pylint
-
-on: [push]
-
-jobs:
-  build:
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        python-version: ["3.8", "3.9", "3.10"]
-    steps:
-    - uses: actions/checkout@v3
-    - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v3
-      with:
-        python-version: ${{ matrix.python-version }}
-    - name: Install dependencies
-      run: |
-        python -m pip install --upgrade pip
-        pip install pylint
-    - name: Analysing the code with pylint
-      run: |
-        pylint $(git ls-files '*.py')

TODO.txt

@@ -1,6 +1,3 @@
-ensembles seem to be broken; they have an internal model selection which takes the parameters, but since quapy now
-    works with protocols it would need to know the validation set in order to pass something like
-    "protocol: APP(val, etc.)"
 sample_size should not be mandatory when qp.environ['SAMPLE_SIZE'] has been specified
 clean all the cumbersome methods that have to be implemented for new quantifiers (e.g., n_classes_ prop, etc.)
 make truly parallel the GridSearchQ

docs/build/html/genindex.html

@@ -224,6 +224,8 @@
       <li><a href="quapy.html#quapy.util.create_parent_dir">create_parent_dir() (in module quapy.util)</a>
 </li>
       <li><a href="quapy.method.html#quapy.method.aggregative.cross_generate_predictions">cross_generate_predictions() (in module quapy.method.aggregative)</a>
+</li>
+      <li><a href="quapy.method.html#quapy.method.aggregative.cross_generate_predictions_depr">cross_generate_predictions_depr() (in module quapy.method.aggregative)</a>
 </li>
       <li><a href="quapy.html#quapy.model_selection.cross_val_predict">cross_val_predict() (in module quapy.model_selection)</a>
 </li>

docs/build/html/quapy.classification.html

@@ -316,14 +316,11 @@ fitting <cite>TruncatedSVD</cite> and then <cite>LogisticRegression</cite> on th
 
 <dl class="py method">
 <dt class="sig sig-object py" id="quapy.classification.methods.LowRankLogisticRegression.get_params">
-<span class="sig-name descname"><span class="pre">get_params</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">deep</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.classification.methods.LowRankLogisticRegression.get_params" title="Permalink to this definition">¶</a></dt>
+<span class="sig-name descname"><span class="pre">get_params</span></span><span class="sig-paren">(</span><span class="sig-paren">)</span><a class="headerlink" href="#quapy.classification.methods.LowRankLogisticRegression.get_params" title="Permalink to this definition">¶</a></dt>
 <dd><p>Get hyper-parameters for this estimator.</p>
 <dl class="field-list simple">
-<dt class="field-odd">Parameters<span class="colon">:</span></dt>
+<dt class="field-odd">Returns<span class="colon">:</span></dt>
-<dd class="field-odd"><p><strong>deep</strong> – compatibility with sklearn</p>
+<dd class="field-odd"><p>a dictionary with parameter names mapped to their values</p>
-</dd>
-<dt class="field-even">Returns<span class="colon">:</span></dt>
-<dd class="field-even"><p>a dictionary with parameter names mapped to their values</p>
 </dd>
 </dl>
 </dd></dl>
@@ -527,7 +524,7 @@ dimensionality of the embedding</p>
 
 <dl class="py class">
 <dt class="sig sig-object py" id="quapy.classification.neural.NeuralClassifierTrainer">
-<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">quapy.classification.neural.</span></span><span class="sig-name descname"><span class="pre">NeuralClassifierTrainer</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">net</span></span><span class="p"><span class="pre">:</span></span><span class="w"> </span><span class="n"><a class="reference internal" href="#quapy.classification.neural.TextClassifierNet" title="quapy.classification.neural.TextClassifierNet"><span class="pre">TextClassifierNet</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">lr</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.001</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">weight_decay</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">patience</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">10</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">epochs</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">200</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">64</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size_test</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">512</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">padding_length</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">300</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'cuda'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">checkpointpath</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'../checkpoint/classifier_net.dat'</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.classification.neural.NeuralClassifierTrainer" title="Permalink to this definition">¶</a></dt>
+<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">quapy.classification.neural.</span></span><span class="sig-name descname"><span class="pre">NeuralClassifierTrainer</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">net</span></span><span class="p"><span class="pre">:</span></span><span class="w"> </span><span class="n"><a class="reference internal" href="#quapy.classification.neural.TextClassifierNet" title="quapy.classification.neural.TextClassifierNet"><span class="pre">TextClassifierNet</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">lr</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0.001</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">weight_decay</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">patience</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">10</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">epochs</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">200</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">64</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">batch_size_test</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">512</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">padding_length</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">300</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">device</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'cpu'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">checkpointpath</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'../checkpoint/classifier_net.dat'</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.classification.neural.NeuralClassifierTrainer" title="Permalink to this definition">¶</a></dt>
 <dd><p>Bases: <code class="xref py py-class docutils literal notranslate"><span class="pre">object</span></code></p>
 <p>Trains a neural network for text classification.</p>
 <dl class="field-list simple">

docs/build/html/quapy.data.html

@@ -447,8 +447,8 @@ index.</p>
 <span class="sig-name descname"><span class="pre">sampling_index</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">size</span></span></em>, <em class="sig-param"><span class="o"><span class="pre">*</span></span><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">shuffle</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">random_state</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.data.base.LabelledCollection.sampling_index" title="Permalink to this definition">¶</a></dt>
 <dd><p>Returns an index to be used to extract a random sample of desired size and desired prevalence values. If the
 prevalence values are not specified, then returns the index of a uniform sampling.
-For each class, the sampling is drawn with replacement if the requested prevalence is larger than
+For each class, the sampling is drawn without replacement if the requested prevalence is larger than
-the actual prevalence of the class, or without replacement otherwise.</p>
+the actual prevalence of the class, or with replacement otherwise.</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
@@ -534,7 +534,7 @@ values for each class)</p>
 <dt class="sig sig-object py" id="quapy.data.base.LabelledCollection.uniform_sampling">
 <span class="sig-name descname"><span class="pre">uniform_sampling</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">size</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">random_state</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.data.base.LabelledCollection.uniform_sampling" title="Permalink to this definition">¶</a></dt>
 <dd><p>Returns a uniform sample (an instance of <a class="reference internal" href="#quapy.data.base.LabelledCollection" title="quapy.data.base.LabelledCollection"><code class="xref py py-class docutils literal notranslate"><span class="pre">LabelledCollection</span></code></a>) of desired size. The sampling is drawn
-with replacement if the requested size is greater than the number of instances, or without replacement
+without replacement if the requested size is greater than the number of instances, or with replacement
 otherwise.</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
@@ -553,7 +553,7 @@ otherwise.</p>
 <dt class="sig sig-object py" id="quapy.data.base.LabelledCollection.uniform_sampling_index">
 <span class="sig-name descname"><span class="pre">uniform_sampling_index</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">size</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">random_state</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.data.base.LabelledCollection.uniform_sampling_index" title="Permalink to this definition">¶</a></dt>
 <dd><p>Returns an index to be used to extract a uniform sample of desired size. The sampling is drawn
-with replacement if the requested size is greater than the number of instances, or without replacement
+without replacement if the requested size is greater than the number of instances, or with replacement
 otherwise.</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>

docs/build/html/quapy.html

@@ -61,7 +61,6 @@
 </section>
 <section id="module-quapy.error">
 <span id="quapy-error"></span><h2>quapy.error<a class="headerlink" href="#module-quapy.error" title="Permalink to this heading">¶</a></h2>
-<p>Implementation of error measures used for quantification</p>
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.absolute_error">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">absolute_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.absolute_error" title="Permalink to this definition">¶</a></dt>
@@ -87,9 +86,8 @@ where <span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.acc_error">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">acc_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">y_true</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">y_pred</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.acc_error" title="Permalink to this definition">¶</a></dt>
-<dd><p>Computes the error in terms of 1-accuracy. The accuracy is computed as
+<dd><p>Computes the error in terms of 1-accuracy. The accuracy is computed as <span class="math notranslate nohighlight">\(\frac{tp+tn}{tp+fp+fn+tn}\)</span>, with
-<span class="math notranslate nohighlight">\(\frac{tp+tn}{tp+fp+fn+tn}\)</span>, with <cite>tp</cite>, <cite>fp</cite>, <cite>fn</cite>, and <cite>tn</cite> standing
+<cite>tp</cite>, <cite>fp</cite>, <cite>fn</cite>, and <cite>tn</cite> standing for true positives, false positives, false negatives, and true negatives,
-for true positives, false positives, false negatives, and true negatives,
 respectively</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
@@ -107,9 +105,8 @@ respectively</p>
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.acce">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">acce</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">y_true</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">y_pred</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.acce" title="Permalink to this definition">¶</a></dt>
-<dd><p>Computes the error in terms of 1-accuracy. The accuracy is computed as
+<dd><p>Computes the error in terms of 1-accuracy. The accuracy is computed as <span class="math notranslate nohighlight">\(\frac{tp+tn}{tp+fp+fn+tn}\)</span>, with
-<span class="math notranslate nohighlight">\(\frac{tp+tn}{tp+fp+fn+tn}\)</span>, with <cite>tp</cite>, <cite>fp</cite>, <cite>fn</cite>, and <cite>tn</cite> standing
+<cite>tp</cite>, <cite>fp</cite>, <cite>fn</cite>, and <cite>tn</cite> standing for true positives, false positives, false negatives, and true negatives,
-for true positives, false positives, false negatives, and true negatives,
 respectively</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
@@ -149,12 +146,10 @@ where <span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.f1_error">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">f1_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">y_true</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">y_pred</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.f1_error" title="Permalink to this definition">¶</a></dt>
-<dd><p>F1 error: simply computes the error in terms of macro <span class="math notranslate nohighlight">\(F_1\)</span>, i.e.,
+<dd><p>F1 error: simply computes the error in terms of macro <span class="math notranslate nohighlight">\(F_1\)</span>, i.e., <span class="math notranslate nohighlight">\(1-F_1^M\)</span>,
-<span class="math notranslate nohighlight">\(1-F_1^M\)</span>, where <span class="math notranslate nohighlight">\(F_1\)</span> is the harmonic mean of precision and recall,
+where <span class="math notranslate nohighlight">\(F_1\)</span> is the harmonic mean of precision and recall, defined as <span class="math notranslate nohighlight">\(\frac{2tp}{2tp+fp+fn}\)</span>,
-defined as <span class="math notranslate nohighlight">\(\frac{2tp}{2tp+fp+fn}\)</span>, with <cite>tp</cite>, <cite>fp</cite>, and <cite>fn</cite> standing
+with <cite>tp</cite>, <cite>fp</cite>, and <cite>fn</cite> standing for true positives, false positives, and false negatives, respectively.
-for true positives, false positives, and false negatives, respectively.
+<cite>Macro</cite> averaging means the <span class="math notranslate nohighlight">\(F_1\)</span> is computed for each category independently, and then averaged.</p>
-<cite>Macro</cite> averaging means the <span class="math notranslate nohighlight">\(F_1\)</span> is computed for each category independently,
-and then averaged.</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
@@ -171,12 +166,10 @@ and then averaged.</p>
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.f1e">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">f1e</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">y_true</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">y_pred</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.f1e" title="Permalink to this definition">¶</a></dt>
-<dd><p>F1 error: simply computes the error in terms of macro <span class="math notranslate nohighlight">\(F_1\)</span>, i.e.,
+<dd><p>F1 error: simply computes the error in terms of macro <span class="math notranslate nohighlight">\(F_1\)</span>, i.e., <span class="math notranslate nohighlight">\(1-F_1^M\)</span>,
-<span class="math notranslate nohighlight">\(1-F_1^M\)</span>, where <span class="math notranslate nohighlight">\(F_1\)</span> is the harmonic mean of precision and recall,
+where <span class="math notranslate nohighlight">\(F_1\)</span> is the harmonic mean of precision and recall, defined as <span class="math notranslate nohighlight">\(\frac{2tp}{2tp+fp+fn}\)</span>,
-defined as <span class="math notranslate nohighlight">\(\frac{2tp}{2tp+fp+fn}\)</span>, with <cite>tp</cite>, <cite>fp</cite>, and <cite>fn</cite> standing
+with <cite>tp</cite>, <cite>fp</cite>, and <cite>fn</cite> standing for true positives, false positives, and false negatives, respectively.
-for true positives, false positives, and false negatives, respectively.
+<cite>Macro</cite> averaging means the <span class="math notranslate nohighlight">\(F_1\)</span> is computed for each category independently, and then averaged.</p>
-<cite>Macro</cite> averaging means the <span class="math notranslate nohighlight">\(F_1\)</span> is computed for each category independently,
-and then averaged.</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
@@ -193,8 +186,7 @@ and then averaged.</p>
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.from_name">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">from_name</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">err_name</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.from_name" title="Permalink to this definition">¶</a></dt>
-<dd><p>Gets an error function from its name. E.g., <cite>from_name(“mae”)</cite>
+<dd><p>Gets an error function from its name. E.g., <cite>from_name(“mae”)</cite> will return function <a class="reference internal" href="#quapy.error.mae" title="quapy.error.mae"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.mae()</span></code></a></p>
-will return function <a class="reference internal" href="#quapy.error.mae" title="quapy.error.mae"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.mae()</span></code></a></p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><p><strong>err_name</strong> – string, the error name</p>
@@ -207,13 +199,11 @@ will return function <a class="reference internal" href="#quapy.error.mae" title
 
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.kld">
-<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">kld</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.kld" title="Permalink to this definition">¶</a></dt>
+<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">kld</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">p</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.kld" title="Permalink to this definition">¶</a></dt>
 <dd><dl class="simple">
-<dt>Computes the Kullback-Leibler divergence between the two prevalence distributions.</dt><dd><p>Kullback-Leibler divergence between two prevalence distributions <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span>
+<dt>Computes the Kullback-Leibler divergence between the two prevalence distributions.</dt><dd><p>Kullback-Leibler divergence between two prevalence distributions <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span> is computed as
-is computed as
+<span class="math notranslate nohighlight">\(KLD(p,\hat{p})=D_{KL}(p||\hat{p})=\sum_{y\in \mathcal{Y}} p(y)\log\frac{p(y)}{\hat{p}(y)}\)</span>, where
-<span class="math notranslate nohighlight">\(KLD(p,\hat{p})=D_{KL}(p||\hat{p})=
+<span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the classes of interest.
-\sum_{y\in \mathcal{Y}} p(y)\log\frac{p(y)}{\hat{p}(y)}\)</span>,
-where <span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the classes of interest.
 The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 </dd>
 </dl>
@@ -222,10 +212,9 @@ The distributions are smoothed using the <cite>eps</cite> factor (see <a class="
 <dd class="field-odd"><ul class="simple">
 <li><p><strong>prevs</strong> – array-like of shape <cite>(n_classes,)</cite> with the true prevalence values</p></li>
 <li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>eps</strong> – smoothing factor. KLD is not defined in cases in which the distributions contain
+<li><p><strong>eps</strong> – smoothing factor. KLD is not defined in cases in which the distributions contain zeros; <cite>eps</cite>
-zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size.
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-If <cite>eps=None</cite>, the sample size will be taken from the environment variable <cite>SAMPLE_SIZE</cite>
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-(which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -242,8 +231,7 @@ If <cite>eps=None</cite>, the sample size will be taken from the environment var
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
 <li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true prevalence values</p></li>
-<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted
+<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted prevalence values</p></li>
-prevalence values</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -260,8 +248,7 @@ prevalence values</p></li>
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
 <li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true prevalence values</p></li>
-<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted
+<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted prevalence values</p></li>
-prevalence values</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -272,21 +259,17 @@ prevalence values</p></li>
 
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.mean_relative_absolute_error">
-<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">mean_relative_absolute_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.mean_relative_absolute_error" title="Permalink to this definition">¶</a></dt>
+<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">mean_relative_absolute_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">p</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.mean_relative_absolute_error" title="Permalink to this definition">¶</a></dt>
-<dd><p>Computes the mean relative absolute error (see <a class="reference internal" href="#quapy.error.rae" title="quapy.error.rae"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.rae()</span></code></a>) across
+<dd><p>Computes the mean relative absolute error (see <a class="reference internal" href="#quapy.error.rae" title="quapy.error.rae"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.rae()</span></code></a>) across the sample pairs.
-the sample pairs. The distributions are smoothed using the <cite>eps</cite> factor (see
+The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
-<a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
-<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true
+<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true prevalence values</p></li>
-prevalence values</p></li>
+<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted
+<li><p><strong>eps</strong> – smoothing factor. <cite>mrae</cite> is not defined in cases in which the true distribution contains zeros; <cite>eps</cite>
-prevalence values</p></li>
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-<li><p><strong>eps</strong> – smoothing factor. <cite>mrae</cite> is not defined in cases in which the true
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-distribution contains zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>,
-with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size will be taken from
-the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -298,20 +281,16 @@ the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set befo
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.mkld">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">mkld</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.mkld" title="Permalink to this definition">¶</a></dt>
-<dd><p>Computes the mean Kullback-Leibler divergence (see <a class="reference internal" href="#quapy.error.kld" title="quapy.error.kld"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.kld()</span></code></a>) across the
+<dd><p>Computes the mean Kullback-Leibler divergence (see <a class="reference internal" href="#quapy.error.kld" title="quapy.error.kld"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.kld()</span></code></a>) across the sample pairs.
-sample pairs. The distributions are smoothed using the <cite>eps</cite> factor
+The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
-(see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
-<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true
+<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true prevalence values</p></li>
-prevalence values</p></li>
+<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted
+<li><p><strong>eps</strong> – smoothing factor. KLD is not defined in cases in which the distributions contain zeros; <cite>eps</cite>
-prevalence values</p></li>
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-<li><p><strong>eps</strong> – smoothing factor. KLD is not defined in cases in which the distributions contain
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size.
-If <cite>eps=None</cite>, the sample size will be taken from the environment variable <cite>SAMPLE_SIZE</cite>
-(which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -323,19 +302,16 @@ If <cite>eps=None</cite>, the sample size will be taken from the environment var
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.mnkld">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">mnkld</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.mnkld" title="Permalink to this definition">¶</a></dt>
-<dd><p>Computes the mean Normalized Kullback-Leibler divergence (see <a class="reference internal" href="#quapy.error.nkld" title="quapy.error.nkld"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.nkld()</span></code></a>)
+<dd><p>Computes the mean Normalized Kullback-Leibler divergence (see <a class="reference internal" href="#quapy.error.nkld" title="quapy.error.nkld"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.nkld()</span></code></a>) across the sample pairs.
-across the sample pairs. The distributions are smoothed using the <cite>eps</cite> factor
+The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
-(see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
 <li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true prevalence values</p></li>
-<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted
+<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted prevalence values</p></li>
-prevalence values</p></li>
+<li><p><strong>eps</strong> – smoothing factor. NKLD is not defined in cases in which the distributions contain zeros; <cite>eps</cite>
-<li><p><strong>eps</strong> – smoothing factor. NKLD is not defined in cases in which the distributions contain
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size.
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-If <cite>eps=None</cite>, the sample size will be taken from the environment variable <cite>SAMPLE_SIZE</cite>
-(which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -346,21 +322,17 @@ If <cite>eps=None</cite>, the sample size will be taken from the environment var
 
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.mrae">
-<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">mrae</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.mrae" title="Permalink to this definition">¶</a></dt>
+<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">mrae</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">p</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.mrae" title="Permalink to this definition">¶</a></dt>
-<dd><p>Computes the mean relative absolute error (see <a class="reference internal" href="#quapy.error.rae" title="quapy.error.rae"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.rae()</span></code></a>) across
+<dd><p>Computes the mean relative absolute error (see <a class="reference internal" href="#quapy.error.rae" title="quapy.error.rae"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.rae()</span></code></a>) across the sample pairs.
-the sample pairs. The distributions are smoothed using the <cite>eps</cite> factor (see
+The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
-<a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
-<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true
+<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true prevalence values</p></li>
-prevalence values</p></li>
+<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted
+<li><p><strong>eps</strong> – smoothing factor. <cite>mrae</cite> is not defined in cases in which the true distribution contains zeros; <cite>eps</cite>
-prevalence values</p></li>
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-<li><p><strong>eps</strong> – smoothing factor. <cite>mrae</cite> is not defined in cases in which the true
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-distribution contains zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>,
-with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size will be taken from
-the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -376,10 +348,8 @@ the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set befo
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
-<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the
+<li><p><strong>prevs</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the true prevalence values</p></li>
-true prevalence values</p></li>
+<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_samples, n_classes,)</cite> with the
-predicted prevalence values</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -390,12 +360,10 @@ predicted prevalence values</p></li>
 
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.nkld">
-<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">nkld</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.nkld" title="Permalink to this definition">¶</a></dt>
+<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">nkld</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">p</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.nkld" title="Permalink to this definition">¶</a></dt>
 <dd><dl class="simple">
-<dt>Computes the Normalized Kullback-Leibler divergence between the two prevalence distributions.</dt><dd><p>Normalized Kullback-Leibler divergence between two prevalence distributions <span class="math notranslate nohighlight">\(p\)</span> and
+<dt>Computes the Normalized Kullback-Leibler divergence between the two prevalence distributions.</dt><dd><p>Normalized Kullback-Leibler divergence between two prevalence distributions <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span>
-<span class="math notranslate nohighlight">\(\hat{p}\)</span> is computed as
+is computed as <span class="math notranslate nohighlight">\(NKLD(p,\hat{p}) = 2\frac{e^{KLD(p,\hat{p})}}{e^{KLD(p,\hat{p})}+1}-1\)</span>, where
-math:<cite>NKLD(p,hat{p}) = 2frac{e^{KLD(p,hat{p})}}{e^{KLD(p,hat{p})}+1}-1</cite>,
-where
 <span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the classes of interest.
 The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 </dd>
@@ -405,10 +373,9 @@ The distributions are smoothed using the <cite>eps</cite> factor (see <a class="
 <dd class="field-odd"><ul class="simple">
 <li><p><strong>prevs</strong> – array-like of shape <cite>(n_classes,)</cite> with the true prevalence values</p></li>
 <li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>eps</strong> – smoothing factor. NKLD is not defined in cases in which the distributions
+<li><p><strong>eps</strong> – smoothing factor. NKLD is not defined in cases in which the distributions contain zeros; <cite>eps</cite>
-contain zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-size. If <cite>eps=None</cite>, the sample size will be taken from the environment variable
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-<cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -419,12 +386,10 @@ size. If <cite>eps=None</cite>, the sample size will be taken from the environme
 
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.rae">
-<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">rae</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.rae" title="Permalink to this definition">¶</a></dt>
+<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">rae</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">p</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.rae" title="Permalink to this definition">¶</a></dt>
 <dd><dl class="simple">
-<dt>Computes the absolute relative error between the two prevalence vectors.</dt><dd><p>Relative absolute error between two prevalence vectors <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span>
+<dt>Computes the absolute relative error between the two prevalence vectors.</dt><dd><p>Relative absolute error between two prevalence vectors <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span>  is computed as
-is computed as
+<span class="math notranslate nohighlight">\(RAE(p,\hat{p})=\frac{1}{|\mathcal{Y}|}\sum_{y\in \mathcal{Y}}\frac{|\hat{p}(y)-p(y)|}{p(y)}\)</span>,
-<span class="math notranslate nohighlight">\(RAE(p,\hat{p})=
-\frac{1}{|\mathcal{Y}|}\sum_{y\in \mathcal{Y}}\frac{|\hat{p}(y)-p(y)|}{p(y)}\)</span>,
 where <span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the classes of interest.
 The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 </dd>
@@ -434,10 +399,9 @@ The distributions are smoothed using the <cite>eps</cite> factor (see <a class="
 <dd class="field-odd"><ul class="simple">
 <li><p><strong>prevs</strong> – array-like of shape <cite>(n_classes,)</cite> with the true prevalence values</p></li>
 <li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>eps</strong> – smoothing factor. <cite>rae</cite> is not defined in cases in which the true distribution
+<li><p><strong>eps</strong> – smoothing factor. <cite>rae</cite> is not defined in cases in which the true distribution contains zeros; <cite>eps</cite>
-contains zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-sample size. If <cite>eps=None</cite>, the sample size will be taken from the environment variable
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-<cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -448,12 +412,10 @@ sample size. If <cite>eps=None</cite>, the sample size will be taken from the en
 
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.relative_absolute_error">
-<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">relative_absolute_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.relative_absolute_error" title="Permalink to this definition">¶</a></dt>
+<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">relative_absolute_error</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">p</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p_hat</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.relative_absolute_error" title="Permalink to this definition">¶</a></dt>
 <dd><dl class="simple">
-<dt>Computes the absolute relative error between the two prevalence vectors.</dt><dd><p>Relative absolute error between two prevalence vectors <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span>
+<dt>Computes the absolute relative error between the two prevalence vectors.</dt><dd><p>Relative absolute error between two prevalence vectors <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span>  is computed as
-is computed as
+<span class="math notranslate nohighlight">\(RAE(p,\hat{p})=\frac{1}{|\mathcal{Y}|}\sum_{y\in \mathcal{Y}}\frac{|\hat{p}(y)-p(y)|}{p(y)}\)</span>,
-<span class="math notranslate nohighlight">\(RAE(p,\hat{p})=
-\frac{1}{|\mathcal{Y}|}\sum_{y\in \mathcal{Y}}\frac{|\hat{p}(y)-p(y)|}{p(y)}\)</span>,
 where <span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the classes of interest.
 The distributions are smoothed using the <cite>eps</cite> factor (see <a class="reference internal" href="#quapy.error.smooth" title="quapy.error.smooth"><code class="xref py py-meth docutils literal notranslate"><span class="pre">quapy.error.smooth()</span></code></a>).</p>
 </dd>
@@ -463,10 +425,9 @@ The distributions are smoothed using the <cite>eps</cite> factor (see <a class="
 <dd class="field-odd"><ul class="simple">
 <li><p><strong>prevs</strong> – array-like of shape <cite>(n_classes,)</cite> with the true prevalence values</p></li>
 <li><p><strong>prevs_hat</strong> – array-like of shape <cite>(n_classes,)</cite> with the predicted prevalence values</p></li>
-<li><p><strong>eps</strong> – smoothing factor. <cite>rae</cite> is not defined in cases in which the true distribution
+<li><p><strong>eps</strong> – smoothing factor. <cite>rae</cite> is not defined in cases in which the true distribution contains zeros; <cite>eps</cite>
-contains zeros; <cite>eps</cite> is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the
+is typically set to be <span class="math notranslate nohighlight">\(\frac{1}{2T}\)</span>, with <span class="math notranslate nohighlight">\(T\)</span> the sample size. If <cite>eps=None</cite>, the sample size
-sample size. If <cite>eps=None</cite>, the sample size will be taken from the environment variable
+will be taken from the environment variable <cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
-<cite>SAMPLE_SIZE</cite> (which has thus to be set beforehand).</p></li>
 </ul>
 </dd>
 <dt class="field-even">Returns<span class="colon">:</span></dt>
@@ -477,11 +438,10 @@ sample size. If <cite>eps=None</cite>, the sample size will be taken from the en
 
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.error.se">
-<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">se</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">prevs_hat</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.se" title="Permalink to this definition">¶</a></dt>
+<span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">se</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">p</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">p_hat</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.se" title="Permalink to this definition">¶</a></dt>
 <dd><dl class="simple">
 <dt>Computes the squared error between the two prevalence vectors.</dt><dd><p>Squared error between two prevalence vectors <span class="math notranslate nohighlight">\(p\)</span> and <span class="math notranslate nohighlight">\(\hat{p}\)</span>  is computed as
-<span class="math notranslate nohighlight">\(SE(p,\hat{p})=\frac{1}{|\mathcal{Y}|}\sum_{y\in \mathcal{Y}}(\hat{p}(y)-p(y))^2\)</span>,
+<span class="math notranslate nohighlight">\(SE(p,\hat{p})=\frac{1}{|\mathcal{Y}|}\sum_{y\in \mathcal{Y}}(\hat{p}(y)-p(y))^2\)</span>, where
-where
 <span class="math notranslate nohighlight">\(\mathcal{Y}\)</span> are the classes of interest.</p>
 </dd>
 </dl>
@@ -502,8 +462,7 @@ where
 <dt class="sig sig-object py" id="quapy.error.smooth">
 <span class="sig-prename descclassname"><span class="pre">quapy.error.</span></span><span class="sig-name descname"><span class="pre">smooth</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">prevs</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">eps</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.error.smooth" title="Permalink to this definition">¶</a></dt>
 <dd><p>Smooths a prevalence distribution with <span class="math notranslate nohighlight">\(\epsilon\)</span> (<cite>eps</cite>) as:
-<span class="math notranslate nohighlight">\(\underline{p}(y)=\frac{\epsilon+p(y)}{\epsilon|\mathcal{Y}|+
+<span class="math notranslate nohighlight">\(\underline{p}(y)=\frac{\epsilon+p(y)}{\epsilon|\mathcal{Y}|+\displaystyle\sum_{y\in \mathcal{Y}}p(y)}\)</span></p>
-\displaystyle\sum_{y\in \mathcal{Y}}p(y)}\)</span></p>
 <dl class="field-list simple">
 <dt class="field-odd">Parameters<span class="colon">:</span></dt>
 <dd class="field-odd"><ul class="simple">
@@ -642,7 +601,7 @@ convenient or not. Set to False to deactivate.</p></li>
 </div>
 <span class="target" id="module-quapy.protocol"></span><dl class="py class">
 <dt class="sig sig-object py" id="quapy.protocol.APP">
-<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">quapy.protocol.</span></span><span class="sig-name descname"><span class="pre">APP</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data</span></span><span class="p"><span class="pre">:</span></span><span class="w"> </span><span class="n"><a class="reference internal" href="quapy.data.html#quapy.data.base.LabelledCollection" title="quapy.data.base.LabelledCollection"><span class="pre">LabelledCollection</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">sample_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">n_prevalences</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">21</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">repeats</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">10</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_limits_epsilon</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">random_state</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">sanity_check</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">10000</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">return_type</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'sample_prev'</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.protocol.APP" title="Permalink to this definition">¶</a></dt>
+<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">quapy.protocol.</span></span><span class="sig-name descname"><span class="pre">APP</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data</span></span><span class="p"><span class="pre">:</span></span><span class="w"> </span><span class="n"><a class="reference internal" href="quapy.data.html#quapy.data.base.LabelledCollection" title="quapy.data.base.LabelledCollection"><span class="pre">LabelledCollection</span></a></span></em>, <em class="sig-param"><span class="n"><span class="pre">sample_size</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">n_prevalences</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">21</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">repeats</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">10</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">smooth_limits_epsilon</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">random_state</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">0</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">return_type</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'sample_prev'</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.protocol.APP" title="Permalink to this definition">¶</a></dt>
 <dd><p>Bases: <a class="reference internal" href="#quapy.protocol.AbstractStochasticSeededProtocol" title="quapy.protocol.AbstractStochasticSeededProtocol"><code class="xref py py-class docutils literal notranslate"><span class="pre">AbstractStochasticSeededProtocol</span></code></a>, <a class="reference internal" href="#quapy.protocol.OnLabelledCollectionProtocol" title="quapy.protocol.OnLabelledCollectionProtocol"><code class="xref py py-class docutils literal notranslate"><span class="pre">OnLabelledCollectionProtocol</span></code></a></p>
 <p>Implementation of the artificial prevalence protocol (APP).
 The APP consists of exploring a grid of prevalence values containing <cite>n_prevalences</cite> points (e.g.,
@@ -662,8 +621,6 @@ grid (default is 21)</p></li>
 <li><p><strong>smooth_limits_epsilon</strong> – the quantity to add and subtract to the limits 0 and 1</p></li>
 <li><p><strong>random_state</strong> – allows replicating samples across runs (default 0, meaning that the sequence of samples
 will be the same every time the protocol is called)</p></li>
-<li><p><strong>sanity_check</strong> – int, raises an exception warning the user that the number of examples to be generated exceed
-this number; set to None for skipping this check</p></li>
 <li><p><strong>return_type</strong> – set to “sample_prev” (default) to get the pairs of (sample, prevalence) at each iteration, or
 to “labelled_collection” to get instead instances of LabelledCollection</p></li>
 </ul>
@@ -1862,7 +1819,6 @@ this function is invoked, it loads the pickled resource. Example:</p>
 </section>
 <section id="module-quapy">
 <span id="module-contents"></span><h2>Module contents<a class="headerlink" href="#module-quapy" title="Permalink to this heading">¶</a></h2>
-<p>QuaPy module for quantification</p>
 </section>
 </section>
 

docs/build/html/quapy.method.html

@@ -1064,6 +1064,11 @@ validation data, or as an integer, indicating that the misclassification rates s
 <span class="sig-prename descclassname"><span class="pre">quapy.method.aggregative.</span></span><span class="sig-name descname"><span class="pre">cross_generate_predictions</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">classifier</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">val_split</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">probabilistic</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">fit_classifier</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">n_jobs</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.method.aggregative.cross_generate_predictions" title="Permalink to this definition">¶</a></dt>
 <dd></dd></dl>
 
+<dl class="py function">
+<dt class="sig sig-object py" id="quapy.method.aggregative.cross_generate_predictions_depr">
+<span class="sig-prename descclassname"><span class="pre">quapy.method.aggregative.</span></span><span class="sig-name descname"><span class="pre">cross_generate_predictions_depr</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">data</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">classifier</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">val_split</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">probabilistic</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">fit_classifier</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">method_name</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">''</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.method.aggregative.cross_generate_predictions_depr" title="Permalink to this definition">¶</a></dt>
+<dd></dd></dl>
+
 <dl class="py function">
 <dt class="sig sig-object py" id="quapy.method.aggregative.newELM">
 <span class="sig-prename descclassname"><span class="pre">quapy.method.aggregative.</span></span><span class="sig-name descname"><span class="pre">newELM</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">svmperf_base</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">loss</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">'01'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">C</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">1</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#quapy.method.aggregative.newELM" title="Permalink to this definition">¶</a></dt>

examples/uci_experiments.py

@@ -1,152 +0,0 @@
-from copy import deepcopy
-
-import quapy as qp
-from sklearn.calibration import CalibratedClassifierCV
-from sklearn.linear_model import LogisticRegression
-from quapy.classification.methods import LowRankLogisticRegression
-from quapy.method.meta import QuaNet
-from quapy.protocol import APP
-from quapy.method.aggregative import CC, ACC, PCC, PACC, MAX, MS, MS2, EMQ, HDy, newSVMAE
-from quapy.method.meta import EHDy
-import numpy as np
-import os
-import pickle
-import itertools
-import argparse
-import torch
-import shutil
-
-
-N_JOBS = -1
-CUDA_N_JOBS = 2
-ENSEMBLE_N_JOBS = -1
-
-qp.environ['SAMPLE_SIZE'] = 100
-
-
-def newLR():
-    return LogisticRegression(max_iter=1000, solver='lbfgs', n_jobs=-1)
-
-
-def calibratedLR():
-    return CalibratedClassifierCV(LogisticRegression(max_iter=1000, solver='lbfgs', n_jobs=-1))
-
-
-__C_range = np.logspace(-3, 3, 7)
-lr_params = {'classifier__C': __C_range, 'classifier__class_weight': [None, 'balanced']}
-svmperf_params = {'classifier__C': __C_range}
-
-
-def quantification_models():
-    yield 'cc', CC(newLR()), lr_params
-    yield 'acc', ACC(newLR()), lr_params
-    yield 'pcc', PCC(newLR()), lr_params
-    yield 'pacc', PACC(newLR()), lr_params
-    yield 'MAX', MAX(newLR()), lr_params
-    yield 'MS', MS(newLR()), lr_params
-    yield 'MS2', MS2(newLR()), lr_params
-    yield 'sldc', EMQ(newLR(), recalib='platt'), lr_params
-    yield 'svmmae', newSVMAE(), svmperf_params
-    yield 'hdy', HDy(newLR()), lr_params
-
-
-def quantification_cuda_models():
-    device = 'cuda' if torch.cuda.is_available() else 'cpu'
-    print(f'Running QuaNet in {device}')
-    learner = LowRankLogisticRegression()
-    yield 'quanet', QuaNet(learner, checkpointdir=args.checkpointdir, device=device), lr_params
-
-
-def evaluate_experiment(true_prevalences, estim_prevalences):
-    print('\nEvaluation Metrics:\n' + '=' * 22)
-    for eval_measure in [qp.error.mae, qp.error.mrae]:
-        err = eval_measure(true_prevalences, estim_prevalences)
-        print(f'\t{eval_measure.__name__}={err:.4f}')
-    print()
-
-
-def result_path(path, dataset_name, model_name, run, optim_loss):
-    return os.path.join(path, f'{dataset_name}-{model_name}-run{run}-{optim_loss}.pkl')
-
-
-def is_already_computed(dataset_name, model_name, run, optim_loss):
-    return os.path.exists(result_path(args.results, dataset_name, model_name, run, optim_loss))
-
-
-def save_results(dataset_name, model_name, run, optim_loss, *results):
-    rpath = result_path(args.results, dataset_name, model_name, run, optim_loss)
-    qp.util.create_parent_dir(rpath)
-    with open(rpath, 'wb') as foo:
-        pickle.dump(tuple(results), foo, pickle.HIGHEST_PROTOCOL)
-
-
-def run(experiment):
-    optim_loss, dataset_name, (model_name, model, hyperparams) = experiment
-    if dataset_name in ['acute.a', 'acute.b', 'iris.1']: return
-
-    collection = qp.datasets.fetch_UCILabelledCollection(dataset_name)
-    for run, data in enumerate(qp.data.Dataset.kFCV(collection, nfolds=5, nrepeats=1)):
-        if is_already_computed(dataset_name, model_name, run=run, optim_loss=optim_loss):
-            print(f'result for dataset={dataset_name} model={model_name} loss={optim_loss} run={run+1}/5 already computed.')
-            continue
-
-        print(f'running dataset={dataset_name} model={model_name} loss={optim_loss} run={run+1}/5')
-        # model selection (hyperparameter optimization for a quantification-oriented loss)
-        train, test = data.train_test
-        train, val = train.split_stratified()
-        if hyperparams is not None:
-            model_selection = qp.model_selection.GridSearchQ(
-                deepcopy(model),
-                param_grid=hyperparams,
-                protocol=APP(val, n_prevalences=21, repeats=25),
-                error=optim_loss,
-                refit=True,
-                timeout=60*60,
-                verbose=True
-            )
-            model_selection.fit(data.training)
-            model = model_selection.best_model()
-            best_params = model_selection.best_params_
-        else:
-            model.fit(data.training)
-            best_params = {}
-
-        # model evaluation
-        true_prevalences, estim_prevalences = qp.evaluation.prediction(
-            model,
-            protocol=APP(test, n_prevalences=21, repeats=100)
-        )
-        test_true_prevalence = data.test.prevalence()
-
-        evaluate_experiment(true_prevalences, estim_prevalences)
-        save_results(dataset_name, model_name, run, optim_loss,
-                     true_prevalences, estim_prevalences,
-                     data.training.prevalence(), test_true_prevalence,
-                     best_params)
-
-
-if __name__ == '__main__':
-    parser = argparse.ArgumentParser(description='Run experiments for Tweeter Sentiment Quantification')
-    parser.add_argument('results', metavar='RESULT_PATH', type=str,
-                        help='path to the directory where to store the results')
-    parser.add_argument('--svmperfpath', metavar='SVMPERF_PATH', type=str, default='../svm_perf_quantification',
-                        help='path to the directory with svmperf')
-    parser.add_argument('--checkpointdir', metavar='PATH', type=str, default='./checkpoint',
-                        help='path to the directory where to dump QuaNet checkpoints')
-    args = parser.parse_args()
-
-    print(f'Result folder: {args.results}')
-    np.random.seed(0)
-
-    qp.environ['SVMPERF_HOME'] = args.svmperfpath
-
-    optim_losses = ['mae']
-    datasets = qp.datasets.UCI_DATASETS
-
-    models = quantification_models()
-    qp.util.parallel(run, itertools.product(optim_losses, datasets, models), n_jobs=N_JOBS)
-
-    models = quantification_cuda_models()
-    qp.util.parallel(run, itertools.product(optim_losses, datasets, models), n_jobs=CUDA_N_JOBS)
-
-    shutil.rmtree(args.checkpointdir, ignore_errors=True)

laboratory/custom_vectorizers.py

@@ -1,244 +0,0 @@
-from scipy.sparse import csc_matrix, csr_matrix
-from sklearn.base import BaseEstimator, TransformerMixin
-from sklearn.feature_extraction.text import TfidfTransformer, TfidfVectorizer, CountVectorizer
-import numpy as np
-from joblib import Parallel, delayed
-import sklearn
-import math
-from scipy.stats import t
-
-
-class ContTable:
-    def __init__(self, tp=0, tn=0, fp=0, fn=0):
-        self.tp=tp
-        self.tn=tn
-        self.fp=fp
-        self.fn=fn
-
-    def get_d(self): return self.tp + self.tn + self.fp + self.fn
-
-    def get_c(self): return self.tp + self.fn
-
-    def get_not_c(self): return self.tn + self.fp
-
-    def get_f(self): return self.tp + self.fp
-
-    def get_not_f(self): return self.tn + self.fn
-
-    def p_c(self): return (1.0*self.get_c())/self.get_d()
-
-    def p_not_c(self): return 1.0-self.p_c()
-
-    def p_f(self): return (1.0*self.get_f())/self.get_d()
-
-    def p_not_f(self): return 1.0-self.p_f()
-
-    def p_tp(self): return (1.0*self.tp) / self.get_d()
-
-    def p_tn(self): return (1.0*self.tn) / self.get_d()
-
-    def p_fp(self): return (1.0*self.fp) / self.get_d()
-
-    def p_fn(self): return (1.0*self.fn) / self.get_d()
-
-    def tpr(self):
-        c = 1.0*self.get_c()
-        return self.tp / c if c > 0.0 else 0.0
-
-    def fpr(self):
-        _c = 1.0*self.get_not_c()
-        return self.fp / _c if _c > 0.0 else 0.0
-
-
-def __ig_factor(p_tc, p_t, p_c):
-    den = p_t * p_c
-    if den != 0.0 and p_tc != 0:
-        return p_tc * math.log(p_tc / den, 2)
-    else:
-        return 0.0
-
-
-def information_gain(cell):
-    return __ig_factor(cell.p_tp(), cell.p_f(), cell.p_c()) + \
-           __ig_factor(cell.p_fp(), cell.p_f(), cell.p_not_c()) +\
-           __ig_factor(cell.p_fn(), cell.p_not_f(), cell.p_c()) + \
-           __ig_factor(cell.p_tn(), cell.p_not_f(), cell.p_not_c())
-
-
-def squared_information_gain(cell):
-    return information_gain(cell)**2
-
-def posneg_information_gain(cell):
-    ig = information_gain(cell)
-    if cell.tpr() < cell.fpr():
-        return -ig
-    else:
-        return ig
-
-def pos_information_gain(cell):
-    if cell.tpr() < cell.fpr():
-        return 0
-    else:
-        return information_gain(cell)
-
-def pointwise_mutual_information(cell):
-    return __ig_factor(cell.p_tp(), cell.p_f(), cell.p_c())
-
-
-def gss(cell):
-    return cell.p_tp()*cell.p_tn() - cell.p_fp()*cell.p_fn()
-
-
-def chi_square(cell):
-    den = cell.p_f() * cell.p_not_f() * cell.p_c() * cell.p_not_c()
-    if den==0.0: return 0.0
-    num = gss(cell)**2
-    return num / den
-
-
-def conf_interval(xt, n):
-    if n>30:
-        z2 = 3.84145882069 # norm.ppf(0.5+0.95/2.0)**2
-    else:
-        z2 = t.ppf(0.5 + 0.95 / 2.0, df=max(n-1,1)) ** 2
-    p = (xt + 0.5 * z2) / (n + z2)
-    amplitude = 0.5 * z2 * math.sqrt((p * (1.0 - p)) / (n + z2))
-    return p, amplitude
-
-
-def strength(minPosRelFreq, minPos, maxNeg):
-    if minPos > maxNeg:
-        return math.log(2.0 * minPosRelFreq, 2.0)
-    else:
-        return 0.0
-
-
-#set cancel_features=True to allow some features to be weighted as 0 (as in the original article)
-#however, for some extremely imbalanced dataset caused all documents to be 0
-def conf_weight(cell, cancel_features=False):
-    c = cell.get_c()
-    not_c = cell.get_not_c()
-    tp = cell.tp
-    fp = cell.fp
-
-    pos_p, pos_amp = conf_interval(tp, c)
-    neg_p, neg_amp = conf_interval(fp, not_c)
-
-    min_pos = pos_p-pos_amp
-    max_neg = neg_p+neg_amp
-    den = (min_pos + max_neg)
-    minpos_relfreq = min_pos / (den if den != 0 else 1)
-
-    str_tplus = strength(minpos_relfreq, min_pos, max_neg);
-
-    if str_tplus == 0 and not cancel_features:
-        return 1e-20
-
-    return str_tplus;
-
-def get_tsr_matrix(cell_matrix, tsr_score_funtion):
-    nC = len(cell_matrix)
-    nF = len(cell_matrix[0])
-    tsr_matrix = [[tsr_score_funtion(cell_matrix[c,f]) for f in range(nF)] for c in range(nC)]
-    return np.array(tsr_matrix)
-
-
-def feature_label_contingency_table(positive_document_indexes, feature_document_indexes, nD):
-    tp_ = len(positive_document_indexes & feature_document_indexes)
-    fp_ = len(feature_document_indexes - positive_document_indexes)
-    fn_ = len(positive_document_indexes - feature_document_indexes)
-    tn_ = nD - (tp_ + fp_ + fn_)
-    return ContTable(tp=tp_, tn=tn_, fp=fp_, fn=fn_)
-
-def category_tables(feature_sets, category_sets, c, nD, nF):
-    return [feature_label_contingency_table(category_sets[c], feature_sets[f], nD) for f in range(nF)]
-
-def get_supervised_matrix(coocurrence_matrix, label_matrix, n_jobs=-1):
-    """
-    Computes the nC x nF supervised matrix M where Mcf is the 4-cell contingency table for feature f and class c.
-    Efficiency O(nF x nC x log(S)) where S is the sparse factor
-    """
-
-    nD, nF = coocurrence_matrix.shape
-    nD2, nC = label_matrix.shape
-
-    if nD != nD2:
-        raise ValueError('Number of rows in coocurrence matrix shape %s and label matrix shape %s is not consistent' %
-                         (coocurrence_matrix.shape,label_matrix.shape))
-
-    def nonzero_set(matrix, col):
-        return set(matrix[:, col].nonzero()[0])
-
-    if isinstance(coocurrence_matrix, csr_matrix):
-        coocurrence_matrix = csc_matrix(coocurrence_matrix)
-    feature_sets = [nonzero_set(coocurrence_matrix, f) for f in range(nF)]
-    category_sets = [nonzero_set(label_matrix, c) for c in range(nC)]
-    cell_matrix = Parallel(n_jobs=n_jobs, backend="threading")(delayed(category_tables)(feature_sets, category_sets, c, nD, nF) for c in range(nC))
-    return np.array(cell_matrix)
-
-
-
-class TSRweighting(BaseEstimator,TransformerMixin):
-    """
-    Supervised Term Weighting function based on any Term Selection Reduction (TSR) function (e.g., information gain,
-    chi-square, etc.) or, more generally, on any function that could be computed on the 4-cell contingency table for
-    each category-feature pair.
-    The supervised_4cell_matrix (a CxF matrix containing the 4-cell contingency tables
-    for each category-feature pair) can be pre-computed (e.g., during the feature selection phase) and passed as an
-    argument.
-    When C>1, i.e., in multiclass scenarios, a global_policy is used in order to determine a single feature-score which
-    informs about its relevance. Accepted policies include "max" (takes the max score across categories), "ave" and "wave"
-    (take the average, or weighted average, across all categories -- weights correspond to the class prevalence), and "sum"
-    (which sums all category scores).
-    """
-
-    def __init__(self, tsr_function, global_policy='max', supervised_4cell_matrix=None, sublinear_tf=True, norm='l2', min_df=3, n_jobs=-1):
-        if global_policy not in ['max', 'ave', 'wave', 'sum']: raise ValueError('Global policy should be in {"max", "ave", "wave", "sum"}')
-        self.tsr_function = tsr_function
-        self.global_policy = global_policy
-        self.supervised_4cell_matrix = supervised_4cell_matrix
-        self.sublinear_tf=sublinear_tf
-        self.norm=norm
-        self.min_df = min_df
-        self.n_jobs=n_jobs
-
-    def fit(self, X, y):
-        self.count_vectorizer = CountVectorizer(min_df=self.min_df)
-        X = self.count_vectorizer.fit_transform(X)
-
-        self.tf_vectorizer = TfidfTransformer(
-            norm=None, use_idf=False, smooth_idf=False, sublinear_tf=self.sublinear_tf).fit(X)
-
-        if len(y.shape) == 1:
-            y = np.expand_dims(y, axis=1)
-
-        nD, nC = y.shape
-        nF = len(self.tf_vectorizer.get_feature_names_out())
-
-        if self.supervised_4cell_matrix is None:
-            self.supervised_4cell_matrix = get_supervised_matrix(X, y, n_jobs=self.n_jobs)
-        else:
-            if self.supervised_4cell_matrix.shape != (nC, nF): raise ValueError("Shape of supervised information matrix is inconsistent with X and y")
-        tsr_matrix = get_tsr_matrix(self.supervised_4cell_matrix, self.tsr_function)
-        if self.global_policy == 'ave':
-            self.global_tsr_vector = np.average(tsr_matrix, axis=0)
-        elif self.global_policy == 'wave':
-            category_prevalences = [sum(y[:,c])*1.0/nD for c in range(nC)]
-            self.global_tsr_vector = np.average(tsr_matrix, axis=0, weights=category_prevalences)
-        elif self.global_policy == 'sum':
-            self.global_tsr_vector = np.sum(tsr_matrix, axis=0)
-        elif self.global_policy == 'max':
-            self.global_tsr_vector = np.amax(tsr_matrix, axis=0)
-        return self
-
-    def fit_transform(self, X, y):
-        return self.fit(X,y).transform(X)
-
-    def transform(self, X):
-        if not hasattr(self, 'global_tsr_vector'): raise NameError('TSRweighting: transform method called before fit.')
-        X = self.count_vectorizer.transform(X)
-        tf_X = self.tf_vectorizer.transform(X).toarray()
-        weighted_X = np.multiply(tf_X, self.global_tsr_vector)
-        if self.norm is not None and self.norm!='none':
-            weighted_X = sklearn.preprocessing.normalize(weighted_X, norm=self.norm, axis=1, copy=False)
-        return csr_matrix(weighted_X)

laboratory/method_dxs.py

@@ -1,148 +0,0 @@
-from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer
-from sklearn.linear_model import LogisticRegression
-
-import quapy as qp
-from data import LabelledCollection
-import numpy as np
-
-from laboratory.custom_vectorizers import *
-from protocol import APP
-from quapy.method.aggregative import _get_divergence, HDy, DistributionMatching
-from quapy.method.base import BaseQuantifier
-from scipy import optimize
-import pandas as pd
-
-
-# TODO: explore the bernoulli (term presence/absence) variant
-# TODO: explore the multinomial (term frequency) variant
-# TODO: explore the multinomial + length normalization variant
-# TODO: consolidate the TSR-variant (e.g., using information gain) variant;
-#   - works better with the idf?
-#   - works better with length normalization?
-#   - etc
-
-class DxS(BaseQuantifier):
-    def __init__(self, vectorizer=None, divergence='topsoe'):
-        self.vectorizer = vectorizer
-        self.divergence = divergence
-
-    # def __as_distribution(self, instances):
-    #     return np.asarray(instances.sum(axis=0) / instances.sum()).flatten()
-
-    def __as_distribution(self, instances):
-        dist = instances.sum(axis=0) / instances.sum()
-        return np.asarray(dist).flatten()
-
-    def fit(self, data: LabelledCollection):
-
-        text_instances, labels = data.Xy
-
-        if self.vectorizer is not None:
-            text_instances = self.vectorizer.fit_transform(text_instances, y=labels)
-
-        distributions = []
-        for class_i in data.classes_:
-            distributions.append(self.__as_distribution(text_instances[labels == class_i]))
-        self.validation_distribution = np.asarray(distributions)
-        return self
-
-    def quantify(self, text_instances):
-        if self.vectorizer is not None:
-            text_instances = self.vectorizer.transform(text_instances)
-
-        test_distribution = self.__as_distribution(text_instances)
-        divergence = _get_divergence(self.divergence)
-        n_classes, n_feats = self.validation_distribution.shape
-
-        def match(prev):
-            prev = np.expand_dims(prev, axis=0)
-            mixture_distribution = (prev @ self.validation_distribution).flatten()
-            return divergence(test_distribution, mixture_distribution)
-
-        # the initial point is set as the uniform distribution
-        uniform_distribution = np.full(fill_value=1 / n_classes, shape=(n_classes,))
-
-        # solutions are bounded to those contained in the unit-simplex
-        bounds = tuple((0, 1) for x in range(n_classes))  # values in [0,1]
-        constraints = ({'type': 'eq', 'fun': lambda x: 1 - sum(x)})  # values summing up to 1
-        r = optimize.minimize(match, x0=uniform_distribution, method='SLSQP', bounds=bounds, constraints=constraints)
-        return r.x
-
-
-
-if __name__ == '__main__':
-
-    qp.environ['SAMPLE_SIZE'] = 250
-    qp.environ['N_JOBS'] = -1
-    min_df = 10
-    # dataset = 'imdb'
-    repeats = 10
-    error = 'mae'
-
-    div = 'HD'
-
-    # generates tuples (dataset, method, method_name)
-    # (the dataset is needed for methods that process the dataset differently)
-    def gen_methods():
-
-        for dataset in qp.datasets.REVIEWS_SENTIMENT_DATASETS:
-
-            data = qp.datasets.fetch_reviews(dataset, tfidf=False)
-
-            bernoulli_vectorizer = CountVectorizer(min_df=min_df, binary=True)
-            dxs = DxS(divergence=div, vectorizer=bernoulli_vectorizer)
-            yield data, dxs, 'DxS-Bernoulli'
-
-            multinomial_vectorizer = CountVectorizer(min_df=min_df, binary=False)
-            dxs = DxS(divergence=div, vectorizer=multinomial_vectorizer)
-            yield data, dxs, 'DxS-multinomial'
-
-            tf_vectorizer = TfidfVectorizer(sublinear_tf=False, use_idf=False, min_df=min_df, norm=None)
-            dxs = DxS(divergence=div, vectorizer=tf_vectorizer)
-            yield data, dxs, 'DxS-TF'
-
-            logtf_vectorizer = TfidfVectorizer(sublinear_tf=True, use_idf=False, min_df=min_df, norm=None)
-            dxs = DxS(divergence=div, vectorizer=logtf_vectorizer)
-            yield data, dxs, 'DxS-logTF'
-
-            tfidf_vectorizer = TfidfVectorizer(use_idf=True, min_df=min_df, norm=None)
-            dxs = DxS(divergence=div, vectorizer=tfidf_vectorizer)
-            yield data, dxs, 'DxS-TFIDF'
-
-            tfidf_vectorizer = TfidfVectorizer(use_idf=True, min_df=min_df, norm='l2')
-            dxs = DxS(divergence=div, vectorizer=tfidf_vectorizer)
-            yield data, dxs, 'DxS-TFIDF-l2'
-
-            tsr_vectorizer = TSRweighting(tsr_function=information_gain, min_df=min_df, norm='l2')
-            dxs = DxS(divergence=div, vectorizer=tsr_vectorizer)
-            yield data, dxs, 'DxS-TFTSR-l2'
-
-            data = qp.datasets.fetch_reviews(dataset, tfidf=True, min_df=min_df)
-            hdy = HDy(LogisticRegression())
-            yield data, hdy, 'HDy'
-
-            dm = DistributionMatching(LogisticRegression(), divergence=div, nbins=5)
-            yield data, dm, 'DM-5b'
-
-            dm = DistributionMatching(LogisticRegression(), divergence=div, nbins=10)
-            yield data, dm, 'DM-10b'
-
-
-    result_path = 'results.csv'
-    with open(result_path, 'wt') as csv:
-        csv.write(f'Method\tDataset\tMAE\tMRAE\n')
-        for data, quantifier, quant_name in gen_methods():
-            quantifier.fit(data.training)
-            report = qp.evaluation.evaluation_report(quantifier, APP(data.test, repeats=repeats), error_metrics=['mae','mrae'], verbose=True)
-            means = report.mean()
-            csv.write(f'{quant_name}\t{data.name}\t{means["mae"]:.5f}\t{means["mrae"]:.5f}\n')
-
-    df = pd.read_csv(result_path, sep='\t')
-    # print(df)
-
-    pv = df.pivot_table(index='Method', columns="Dataset", values=["MAE", "MRAE"])
-    print(pv)
-
-
-
-

laboratory/method_kdey.py

@@ -1,168 +0,0 @@
-from typing import Union, Callable
-import numpy as np
-from sklearn.base import BaseEstimator
-from sklearn.linear_model import LogisticRegression
-import pandas as pd
-from sklearn.neighbors import KernelDensity
-
-import quapy as qp
-from data import LabelledCollection
-from protocol import APP, UPP
-from quapy.method.aggregative import AggregativeProbabilisticQuantifier, _training_helper, cross_generate_predictions, \
-    DistributionMatching, _get_divergence
-import scipy
-from scipy import optimize
-
-
-class KDEy(AggregativeProbabilisticQuantifier):
-
-    BANDWIDTH_METHOD = ['auto', 'scott', 'silverman']
-    ENGINE = ['scipy', 'sklearn']
-
-    def __init__(self, classifier: BaseEstimator, val_split=0.4, divergence: Union[str, Callable]='HD',
-                 bandwidth_method='scott', engine='sklearn', n_jobs=None):
-        self.classifier = classifier
-        self.val_split = val_split
-        self.divergence = divergence
-        self.bandwidth_method = bandwidth_method
-        self.engine = engine
-        self.n_jobs = n_jobs
-        assert bandwidth_method in KDEy.BANDWIDTH_METHOD, f'unknown bandwidth_method, valid ones are {KDEy.BANDWIDTH_METHOD}'
-        assert engine in KDEy.ENGINE, f'unknown engine, valid ones are {KDEy.ENGINE}'
-
-    def get_kde(self, posteriors):
-        if self.engine == 'scipy':
-            # scipy treats columns as datapoints, and need the datapoints not to lie in a lower-dimensional subspace, which
-            # requires removing the last dimension which is constrained
-            posteriors = posteriors[:,:-1].T
-            kde = scipy.stats.gaussian_kde(posteriors)
-            kde.set_bandwidth(self.bandwidth_method)
-        elif self.engine == 'sklearn':
-            kde = KernelDensity(bandwidth=self.bandwidth_method).fit(posteriors)
-        return kde
-
-    def pdf(self, kde, posteriors):
-        if self.engine == 'scipy':
-            return kde(posteriors[:,:-1].T)
-        elif self.engine == 'sklearn':
-            return np.exp(kde.score_samples(posteriors))
-
-    def fit(self, data: LabelledCollection, fit_classifier=True, val_split: Union[float, LabelledCollection] = None):
-        """
-        Trains the classifier (if requested) and generates the validation distributions out of the training data.
-        The validation distributions have shape `(n, ch, nbins)`, with `n` the number of classes, `ch` the number of
-        channels (a channel is a description, in form of a histogram, of a specific class -- there are as many channels
-        as classes, although in the binary case one can use only one channel, since the other one is constrained),
-        and `nbins` the number of bins. In particular, let `V` be the validation distributions; `di=V[i]`
-        are the distributions obtained from training data labelled with class `i`; `dij = di[j]` is the discrete
-        distribution of posterior probabilities `P(Y=j|X=x)` for training data labelled with class `i`, and `dij[k]`
-        is the fraction of instances with a value in the `k`-th bin.
-
-        :param data: the training set
-        :param fit_classifier: set to False to bypass the training (the learner is assumed to be already fit)
-        :param val_split: either a float in (0,1) indicating the proportion of training instances to use for
-         validation (e.g., 0.3 for using 30% of the training set as validation data), or a LabelledCollection
-         indicating the validation set itself, or an int indicating the number k of folds to be used in kFCV
-         to estimate the parameters
-        """
-        if val_split is None:
-            val_split = self.val_split
-
-        self.classifier, y, posteriors, classes, class_count = cross_generate_predictions(
-            data, self.classifier, val_split, probabilistic=True, fit_classifier=fit_classifier, n_jobs=self.n_jobs
-        )
-
-        self.val_densities = [self.get_kde(posteriors[y == cat]) for cat in range(data.n_classes)]
-        self.val_posteriors = posteriors
-
-        return self
-
-    def val_pdf(self, prev):
-        """
-        Returns a function that computes the mixture model with the given prev as mixture factor
-        :param prev: a prevalence vector, ndarray
-        :return: a function implementing the validation distribution with fixed mixture factor
-        """
-        return lambda posteriors: sum(prev_i * self.pdf(kde_i, posteriors) for kde_i, prev_i in zip(self.val_densities, prev))
-
-
-    def aggregate(self, posteriors: np.ndarray):
-        """
-        Searches for the mixture model parameter (the sought prevalence values) that yields a validation distribution
-        (the mixture) that best matches the test distribution, in terms of the divergence measure of choice.
-        In the multiclass case, with `n` the number of classes, the test and mixture distributions contain
-        `n` channels (proper distributions of binned posterior probabilities), on which the divergence is computed
-        independently. The matching is computed as an average of the divergence across all channels.
-
-        :param instances: instances in the sample
-        :return: a vector of class prevalence estimates
-        """
-        test_density = self.get_kde(posteriors)
-        # val_test_posteriors = np.concatenate([self.val_posteriors, posteriors])
-        test_likelihood = self.pdf(test_density, posteriors)
-        divergence = _get_divergence(self.divergence)
-
-
-        n_classes = len(self.val_densities)
-
-        def match(prev):
-            val_pdf = self.val_pdf(prev)
-            val_likelihood  = val_pdf(posteriors)
-            return divergence(val_likelihood, test_likelihood)
-
-        # the initial point is set as the uniform distribution
-        uniform_distribution = np.full(fill_value=1 / n_classes, shape=(n_classes,))
-
-        # solutions are bounded to those contained in the unit-simplex
-        bounds = tuple((0, 1) for _ in range(n_classes))  # values in [0,1]
-        constraints = ({'type': 'eq', 'fun': lambda x: 1 - sum(x)})  # values summing up to 1
-        r = optimize.minimize(match, x0=uniform_distribution, method='SLSQP', bounds=bounds, constraints=constraints)
-        return r.x
-
-
-
-if __name__ == '__main__':
-
-    qp.environ['SAMPLE_SIZE'] = 100
-    qp.environ['N_JOBS'] = -1
-    div = 'HD'
-
-    # generates tuples (dataset, method, method_name)
-    # (the dataset is needed for methods that process the dataset differently)
-    def gen_methods():
-
-        for dataset in qp.datasets.TWITTER_SENTIMENT_DATASETS_TEST:
-
-            data = qp.datasets.fetch_twitter(dataset, min_df=3, pickle=True)
-
-            # kdey = KDEy(LogisticRegression(), divergence=div, bandwidth_method='scott')
-            # yield data, kdey, f'KDEy-{div}-scott'
-
-            kdey = KDEy(LogisticRegression(), divergence=div, bandwidth_method='silverman', engine='sklearn')
-            yield data, kdey, f'KDEy-{div}-silverman'
-
-            dm = DistributionMatching(LogisticRegression(), divergence=div, nbins=5)
-            yield data, dm, f'DM-5b-{div}'
-
-            # dm = DistributionMatching(LogisticRegression(), divergence=div, nbins=10)
-            # yield data, dm, f'DM-10b-{div}'
-
-
-    result_path = 'results_kdey.csv'
-    with open(result_path, 'wt') as csv:
-        csv.write(f'Method\tDataset\tMAE\tMRAE\n')
-        for data, quantifier, quant_name in gen_methods():
-            quantifier.fit(data.training)
-            protocol = UPP(data.test, repeats=100)
-            report = qp.evaluation.evaluation_report(quantifier, protocol, error_metrics=['mae','mrae'], verbose=True)
-            means = report.mean()
-            csv.write(f'{quant_name}\t{data.name}\t{means["mae"]:.5f}\t{means["mrae"]:.5f}\n')
-            csv.flush()
-
-    df = pd.read_csv(result_path, sep='\t')
-    # print(df)
-
-    pd.set_option('display.max_columns', None)
-    pd.set_option('display.max_rows', None)
-    pv = df.pivot_table(index='Dataset', columns="Method", values=["MAE", "MRAE"])
-    print(pv)

quapy/__init__.py

@@ -1,7 +1,6 @@
-"""QuaPy module for quantification"""
-from quapy.data import datasets
 from . import error
 from . import data
+from quapy.data import datasets
 from . import functional
 # from . import method
 from . import evaluation
@@ -26,8 +25,7 @@ environ = {
 
 def _get_njobs(n_jobs):
     """
-    If `n_jobs` is None, then it returns `environ['N_JOBS']`;
+    If `n_jobs` is None, then it returns `environ['N_JOBS']`; if otherwise, returns `n_jobs`.
-    if otherwise, returns `n_jobs`.
 
     :param n_jobs: the number of `n_jobs` or None if not specified
     :return: int
@@ -37,8 +35,7 @@ def _get_njobs(n_jobs):
 
 def _get_sample_size(sample_size):
     """
-    If `sample_size` is None, then it returns `environ['SAMPLE_SIZE']`;
+    If `sample_size` is None, then it returns `environ['SAMPLE_SIZE']`; if otherwise, returns `sample_size`.
-    if otherwise, returns `sample_size`.
     If none of these are set, then a ValueError exception is raised.
 
     :param sample_size: integer or None
@@ -48,3 +45,6 @@ def _get_sample_size(sample_size):
     if sample_size is None:
         raise ValueError('neither sample_size nor qp.environ["SAMPLE_SIZE"] have been specified')
     return sample_size
+
+
+

quapy/classification/methods.py

@@ -19,7 +19,7 @@ class LowRankLogisticRegression(BaseEstimator):
 
     def __init__(self, n_components=100, **kwargs):
         self.n_components = n_components
-        self.classifier = LogisticRegression(**kwargs)
+        self.learner = LogisticRegression(**kwargs)
 
     def get_params(self):
         """
@@ -28,7 +28,7 @@ class LowRankLogisticRegression(BaseEstimator):
         :return: a dictionary with parameter names mapped to their values
         """
         params = {'n_components': self.n_components}
-        params.update(self.classifier.get_params())
+        params.update(self.learner.get_params())
         return params
 
     def set_params(self, **params):
@@ -43,7 +43,7 @@ class LowRankLogisticRegression(BaseEstimator):
         if 'n_components' in params_:
             self.n_components = params_['n_components']
             del params_['n_components']
-        self.classifier.set_params(**params_)
+        self.learner.set_params(**params_)
 
     def fit(self, X, y):
         """
@@ -59,8 +59,8 @@ class LowRankLogisticRegression(BaseEstimator):
         if nF > self.n_components:
             self.pca = TruncatedSVD(self.n_components).fit(X)
         X = self.transform(X)
-        self.classifier.fit(X, y)
+        self.learner.fit(X, y)
-        self.classes_ = self.classifier.classes_
+        self.classes_ = self.learner.classes_
         return self
 
     def predict(self, X):
@@ -72,7 +72,7 @@ class LowRankLogisticRegression(BaseEstimator):
             instances in `X`
         """
         X = self.transform(X)
-        return self.classifier.predict(X)
+        return self.learner.predict(X)
 
     def predict_proba(self, X):
         """
@@ -82,7 +82,7 @@ class LowRankLogisticRegression(BaseEstimator):
         :return: array-like of shape `(n_samples, n_classes)` with the posterior probabilities
         """
         X = self.transform(X)
-        return self.classifier.predict_proba(X)
+        return self.learner.predict_proba(X)
 
     def transform(self, X):
         """

quapy/data/datasets.py

@@ -207,7 +207,7 @@ def fetch_UCIDataset(dataset_name, data_home=None, test_split=0.3, verbose=False
     return Dataset(*data.split_stratified(1 - test_split, random_state=0))
 
 
-def fetch_UCILabelledCollection(dataset_name, data_home=None, verbose=False) -> LabelledCollection:
+def fetch_UCILabelledCollection(dataset_name, data_home=None, verbose=False) -> Dataset:
     """
     Loads a UCI collection as an instance of :class:`quapy.data.base.LabelledCollection`, as used in
     `Pérez-Gállego, P., Quevedo, J. R., & del Coz, J. J. (2017).
@@ -223,7 +223,7 @@ def fetch_UCILabelledCollection(dataset_name, data_home=None, verbose=False) ->
 
     >>> import quapy as qp
     >>> collection = qp.datasets.fetch_UCILabelledCollection("yeast")
-    >>> for data in qp.domains.Dataset.kFCV(collection, nfolds=5, nrepeats=2):
+    >>> for data in qp.data.Dataset.kFCV(collection, nfolds=5, nrepeats=2):
     >>>     ...
 
     The list of valid dataset names can be accessed in `quapy.data.datasets.UCI_DATASETS`
@@ -233,7 +233,7 @@ def fetch_UCILabelledCollection(dataset_name, data_home=None, verbose=False) ->
         ~/quay_data/ directory)
     :param test_split: proportion of documents to be included in the test set. The rest conforms the training set
     :param verbose: set to True (default is False) to get information (from the UCI ML repository) about the datasets
-    :return: a :class:`quapy.data.base.LabelledCollection` instance
+    :return: a :class:`quapy.data.base.Dataset` instance
     """
 
     assert dataset_name in UCI_DATASETS, \

quapy/error.py

@@ -1,13 +1,10 @@
-"""Implementation of error measures used for quantification"""
+import quapy as qp
-
 import numpy as np
 from sklearn.metrics import f1_score
-import quapy as qp
 
 
 def from_name(err_name):
-    """Gets an error function from its name. E.g., `from_name("mae")`
+    """Gets an error function from its name. E.g., `from_name("mae")` will return function :meth:`quapy.error.mae`
-    will return function :meth:`quapy.error.mae`
 
     :param err_name: string, the error name
     :return: a callable implementing the requested error
@@ -18,12 +15,10 @@ def from_name(err_name):
 
 
 def f1e(y_true, y_pred):
-    """F1 error: simply computes the error in terms of macro :math:`F_1`, i.e.,
+    """F1 error: simply computes the error in terms of macro :math:`F_1`, i.e., :math:`1-F_1^M`,
-    :math:`1-F_1^M`, where :math:`F_1` is the harmonic mean of precision and recall,
+    where :math:`F_1` is the harmonic mean of precision and recall, defined as :math:`\\frac{2tp}{2tp+fp+fn}`,
-    defined as :math:`\\frac{2tp}{2tp+fp+fn}`, with `tp`, `fp`, and `fn` standing
+    with `tp`, `fp`, and `fn` standing for true positives, false positives, and false negatives, respectively.
-    for true positives, false positives, and false negatives, respectively.
+    `Macro` averaging means the :math:`F_1` is computed for each category independently, and then averaged.
-    `Macro` averaging means the :math:`F_1` is computed for each category independently,
-    and then averaged.
 
     :param y_true: array-like of true labels
     :param y_pred: array-like of predicted labels
@@ -33,9 +28,8 @@ def f1e(y_true, y_pred):
 
 
 def acce(y_true, y_pred):
-    """Computes the error in terms of 1-accuracy. The accuracy is computed as
+    """Computes the error in terms of 1-accuracy. The accuracy is computed as :math:`\\frac{tp+tn}{tp+fp+fn+tn}`, with
-    :math:`\\frac{tp+tn}{tp+fp+fn+tn}`, with `tp`, `fp`, `fn`, and `tn` standing
+    `tp`, `fp`, `fn`, and `tn` standing for true positives, false positives, false negatives, and true negatives,
-    for true positives, false positives, false negatives, and true negatives,
     respectively
 
     :param y_true: array-like of true labels
@@ -49,8 +43,7 @@ def mae(prevs, prevs_hat):
     """Computes the mean absolute error (see :meth:`quapy.error.ae`) across the sample pairs.
 
     :param prevs: array-like of shape `(n_samples, n_classes,)` with the true prevalence values
-    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted
+    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted prevalence values
-        prevalence values
     :return: mean absolute error
     """
     return ae(prevs, prevs_hat).mean()
@@ -59,7 +52,7 @@ def mae(prevs, prevs_hat):
 def ae(prevs, prevs_hat):
     """Computes the absolute error between the two prevalence vectors.
      Absolute error between two prevalence vectors :math:`p` and :math:`\\hat{p}`  is computed as
-     :math:`AE(p,\\hat{p})=\\frac{1}{|\\mathcal{Y}|}\\sum_{y\\in \\mathcal{Y}}|\\hat{p}(y)-p(y)|`,
+     :math:`AE(p,\\hat{p})=\\frac{1}{|\\mathcal{Y}|}\\sum_{y\in \mathcal{Y}}|\\hat{p}(y)-p(y)|`,
      where :math:`\\mathcal{Y}` are the classes of interest.
 
     :param prevs: array-like of shape `(n_classes,)` with the true prevalence values
@@ -73,153 +66,129 @@ def ae(prevs, prevs_hat):
 def mse(prevs, prevs_hat):
     """Computes the mean squared error (see :meth:`quapy.error.se`) across the sample pairs.
 
-    :param prevs: array-like of shape `(n_samples, n_classes,)` with the
+    :param prevs: array-like of shape `(n_samples, n_classes,)` with the true prevalence values
-        true prevalence values
+    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted prevalence values
-    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the
-        predicted prevalence values
     :return: mean squared error
     """
     return se(prevs, prevs_hat).mean()
 
 
-def se(prevs, prevs_hat):
+def se(p, p_hat):
     """Computes the squared error between the two prevalence vectors.
      Squared error between two prevalence vectors :math:`p` and :math:`\\hat{p}`  is computed as
-     :math:`SE(p,\\hat{p})=\\frac{1}{|\\mathcal{Y}|}\\sum_{y\\in \\mathcal{Y}}(\\hat{p}(y)-p(y))^2`,
+     :math:`SE(p,\\hat{p})=\\frac{1}{|\\mathcal{Y}|}\\sum_{y\in \mathcal{Y}}(\\hat{p}(y)-p(y))^2`, where
-     where
      :math:`\\mathcal{Y}` are the classes of interest.
 
     :param prevs: array-like of shape `(n_classes,)` with the true prevalence values
     :param prevs_hat: array-like of shape `(n_classes,)` with the predicted prevalence values
     :return: absolute error
     """
-    return ((prevs_hat - prevs) ** 2).mean(axis=-1)
+    return ((p_hat-p)**2).mean(axis=-1)
 
 
 def mkld(prevs, prevs_hat, eps=None):
-    """Computes the mean Kullback-Leibler divergence (see :meth:`quapy.error.kld`) across the
+    """Computes the mean Kullback-Leibler divergence (see :meth:`quapy.error.kld`) across the sample pairs.
-    sample pairs. The distributions are smoothed using the `eps` factor
+    The distributions are smoothed using the `eps` factor (see :meth:`quapy.error.smooth`).
-    (see :meth:`quapy.error.smooth`).
 
-    :param prevs: array-like of shape `(n_samples, n_classes,)` with the true
+    :param prevs: array-like of shape `(n_samples, n_classes,)` with the true prevalence values
-        prevalence values
+    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted prevalence values
-    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted
+    :param eps: smoothing factor. KLD is not defined in cases in which the distributions contain zeros; `eps`
-        prevalence values
+        is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size. If `eps=None`, the sample size
-    :param eps: smoothing factor. KLD is not defined in cases in which the distributions contain
+        will be taken from the environment variable `SAMPLE_SIZE` (which has thus to be set beforehand).
-        zeros; `eps` is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size.
-        If `eps=None`, the sample size will be taken from the environment variable `SAMPLE_SIZE`
-        (which has thus to be set beforehand).
     :return: mean Kullback-Leibler distribution
     """
     return kld(prevs, prevs_hat, eps).mean()
 
 
-def kld(prevs, prevs_hat, eps=None):
+def kld(p, p_hat, eps=None):
     """Computes the Kullback-Leibler divergence between the two prevalence distributions.
-     Kullback-Leibler divergence between two prevalence distributions :math:`p` and :math:`\\hat{p}`
+     Kullback-Leibler divergence between two prevalence distributions :math:`p` and :math:`\\hat{p}` is computed as
-     is computed as
+     :math:`KLD(p,\\hat{p})=D_{KL}(p||\\hat{p})=\\sum_{y\\in \\mathcal{Y}} p(y)\\log\\frac{p(y)}{\\hat{p}(y)}`, where
-     :math:`KLD(p,\\hat{p})=D_{KL}(p||\\hat{p})=
-     \\sum_{y\\in \\mathcal{Y}} p(y)\\log\\frac{p(y)}{\\hat{p}(y)}`,
-     where :math:`\\mathcal{Y}` are the classes of interest.
-     The distributions are smoothed using the `eps` factor (see :meth:`quapy.error.smooth`).
-
-    :param prevs: array-like of shape `(n_classes,)` with the true prevalence values
-    :param prevs_hat: array-like of shape `(n_classes,)` with the predicted prevalence values
-    :param eps: smoothing factor. KLD is not defined in cases in which the distributions contain
-        zeros; `eps` is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size.
-        If `eps=None`, the sample size will be taken from the environment variable `SAMPLE_SIZE`
-        (which has thus to be set beforehand).
-    :return: Kullback-Leibler divergence between the two distributions
-    """
-    eps = __check_eps(eps)
-    smooth_prevs = prevs + eps
-    smooth_prevs_hat = prevs_hat + eps
-    return (smooth_prevs*np.log(smooth_prevs/smooth_prevs_hat)).sum(axis=-1)
-
-
-def mnkld(prevs, prevs_hat, eps=None):
-    """Computes the mean Normalized Kullback-Leibler divergence (see :meth:`quapy.error.nkld`)
-    across the sample pairs. The distributions are smoothed using the `eps` factor
-    (see :meth:`quapy.error.smooth`).
-
-    :param prevs: array-like of shape `(n_samples, n_classes,)` with the true prevalence values
-    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted
-        prevalence values
-    :param eps: smoothing factor. NKLD is not defined in cases in which the distributions contain
-        zeros; `eps` is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size.
-        If `eps=None`, the sample size will be taken from the environment variable `SAMPLE_SIZE`
-        (which has thus to be set beforehand).
-    :return: mean Normalized Kullback-Leibler distribution
-    """
-    return nkld(prevs, prevs_hat, eps).mean()
-
-
-def nkld(prevs, prevs_hat, eps=None):
-    """Computes the Normalized Kullback-Leibler divergence between the two prevalence distributions.
-     Normalized Kullback-Leibler divergence between two prevalence distributions :math:`p` and
-     :math:`\\hat{p}` is computed as
-     math:`NKLD(p,\\hat{p}) = 2\\frac{e^{KLD(p,\\hat{p})}}{e^{KLD(p,\\hat{p})}+1}-1`,
-     where
      :math:`\\mathcal{Y}` are the classes of interest.
      The distributions are smoothed using the `eps` factor (see :meth:`quapy.error.smooth`).
 
     :param prevs: array-like of shape `(n_classes,)` with the true prevalence values
     :param prevs_hat: array-like of shape `(n_classes,)` with the predicted prevalence values
-    :param eps: smoothing factor. NKLD is not defined in cases in which the distributions
+    :param eps: smoothing factor. KLD is not defined in cases in which the distributions contain zeros; `eps`
-        contain zeros; `eps` is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample
+        is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size. If `eps=None`, the sample size
-        size. If `eps=None`, the sample size will be taken from the environment variable
+        will be taken from the environment variable `SAMPLE_SIZE` (which has thus to be set beforehand).
-        `SAMPLE_SIZE` (which has thus to be set beforehand).
+    :return: Kullback-Leibler divergence between the two distributions
+    """
+    eps = __check_eps(eps)
+    sp = p+eps
+    sp_hat = p_hat + eps
+    return (sp*np.log(sp/sp_hat)).sum(axis=-1)
+
+
+def mnkld(prevs, prevs_hat, eps=None):
+    """Computes the mean Normalized Kullback-Leibler divergence (see :meth:`quapy.error.nkld`) across the sample pairs.
+    The distributions are smoothed using the `eps` factor (see :meth:`quapy.error.smooth`).
+
+    :param prevs: array-like of shape `(n_samples, n_classes,)` with the true prevalence values
+    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted prevalence values
+    :param eps: smoothing factor. NKLD is not defined in cases in which the distributions contain zeros; `eps`
+        is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size. If `eps=None`, the sample size
+        will be taken from the environment variable `SAMPLE_SIZE` (which has thus to be set beforehand).
+    :return: mean Normalized Kullback-Leibler distribution
+    """
+    return nkld(prevs, prevs_hat, eps).mean()
+
+
+def nkld(p, p_hat, eps=None):
+    """Computes the Normalized Kullback-Leibler divergence between the two prevalence distributions.
+     Normalized Kullback-Leibler divergence between two prevalence distributions :math:`p` and :math:`\\hat{p}`
+     is computed as :math:`NKLD(p,\\hat{p}) = 2\\frac{e^{KLD(p,\\hat{p})}}{e^{KLD(p,\\hat{p})}+1}-1`, where
+     :math:`\\mathcal{Y}` are the classes of interest.
+     The distributions are smoothed using the `eps` factor (see :meth:`quapy.error.smooth`).
+
+    :param prevs: array-like of shape `(n_classes,)` with the true prevalence values
+    :param prevs_hat: array-like of shape `(n_classes,)` with the predicted prevalence values
+    :param eps: smoothing factor. NKLD is not defined in cases in which the distributions contain zeros; `eps`
+        is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size. If `eps=None`, the sample size
+        will be taken from the environment variable `SAMPLE_SIZE` (which has thus to be set beforehand).
     :return: Normalized Kullback-Leibler divergence between the two distributions
     """
-    ekld = np.exp(kld(prevs, prevs_hat, eps))
+    ekld = np.exp(kld(p, p_hat, eps))
     return 2. * ekld / (1 + ekld) - 1.
 
 
-def mrae(prevs, prevs_hat, eps=None):
+def mrae(p, p_hat, eps=None):
-    """Computes the mean relative absolute error (see :meth:`quapy.error.rae`) across
+    """Computes the mean relative absolute error (see :meth:`quapy.error.rae`) across the sample pairs.
-    the sample pairs. The distributions are smoothed using the `eps` factor (see
+    The distributions are smoothed using the `eps` factor (see :meth:`quapy.error.smooth`).
-    :meth:`quapy.error.smooth`).
 
-    :param prevs: array-like of shape `(n_samples, n_classes,)` with the true
+    :param prevs: array-like of shape `(n_samples, n_classes,)` with the true prevalence values
-        prevalence values
+    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted prevalence values
-    :param prevs_hat: array-like of shape `(n_samples, n_classes,)` with the predicted
+    :param eps: smoothing factor. `mrae` is not defined in cases in which the true distribution contains zeros; `eps`
-        prevalence values
+        is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size. If `eps=None`, the sample size
-    :param eps: smoothing factor. `mrae` is not defined in cases in which the true
+        will be taken from the environment variable `SAMPLE_SIZE` (which has thus to be set beforehand).
-        distribution contains zeros; `eps` is typically set to be :math:`\\frac{1}{2T}`,
-        with :math:`T` the sample size. If `eps=None`, the sample size will be taken from
-        the environment variable `SAMPLE_SIZE` (which has thus to be set beforehand).
     :return: mean relative absolute error
     """
-    return rae(prevs, prevs_hat, eps).mean()
+    return rae(p, p_hat, eps).mean()
 
 
-def rae(prevs, prevs_hat, eps=None):
+def rae(p, p_hat, eps=None):
     """Computes the absolute relative error between the two prevalence vectors.
-     Relative absolute error between two prevalence vectors :math:`p` and :math:`\\hat{p}`
+     Relative absolute error between two prevalence vectors :math:`p` and :math:`\\hat{p}`  is computed as
-     is computed as
+     :math:`RAE(p,\\hat{p})=\\frac{1}{|\\mathcal{Y}|}\\sum_{y\in \mathcal{Y}}\\frac{|\\hat{p}(y)-p(y)|}{p(y)}`,
-     :math:`RAE(p,\\hat{p})=
-     \\frac{1}{|\\mathcal{Y}|}\\sum_{y\\in \\mathcal{Y}}\\frac{|\\hat{p}(y)-p(y)|}{p(y)}`,
      where :math:`\\mathcal{Y}` are the classes of interest.
      The distributions are smoothed using the `eps` factor (see :meth:`quapy.error.smooth`).
 
     :param prevs: array-like of shape `(n_classes,)` with the true prevalence values
     :param prevs_hat: array-like of shape `(n_classes,)` with the predicted prevalence values
-    :param eps: smoothing factor. `rae` is not defined in cases in which the true distribution
+    :param eps: smoothing factor. `rae` is not defined in cases in which the true distribution contains zeros; `eps`
-        contains zeros; `eps` is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the
+        is typically set to be :math:`\\frac{1}{2T}`, with :math:`T` the sample size. If `eps=None`, the sample size
-        sample size. If `eps=None`, the sample size will be taken from the environment variable
+        will be taken from the environment variable `SAMPLE_SIZE` (which has thus to be set beforehand).
-        `SAMPLE_SIZE` (which has thus to be set beforehand).
     :return: relative absolute error
     """
     eps = __check_eps(eps)
-    prevs = smooth(prevs, eps)
+    p = smooth(p, eps)
-    prevs_hat = smooth(prevs_hat, eps)
+    p_hat = smooth(p_hat, eps)
-    return (abs(prevs - prevs_hat) / prevs).mean(axis=-1)
+    return (abs(p-p_hat)/p).mean(axis=-1)
 
 
 def smooth(prevs, eps):
-    """ Smooths a prevalence distribution with :math:`\\epsilon` (`eps`) as:
+    """ Smooths a prevalence distribution with :math:`\epsilon` (`eps`) as:
-    :math:`\\underline{p}(y)=\\frac{\\epsilon+p(y)}{\\epsilon|\\mathcal{Y}|+
+    :math:`\\underline{p}(y)=\\frac{\\epsilon+p(y)}{\\epsilon|\\mathcal{Y}|+\\displaystyle\\sum_{y\\in \\mathcal{Y}}p(y)}`
-    \\displaystyle\\sum_{y\\in \\mathcal{Y}}p(y)}`
 
     :param prevs: array-like of shape `(n_classes,)` with the true prevalence values
     :param eps: smoothing factor
@@ -231,10 +200,12 @@ def smooth(prevs, eps):
 
 def __check_eps(eps=None):
     if eps is None:
+        import quapy as qp
         sample_size = qp.environ['SAMPLE_SIZE']
         if sample_size is None:
             raise ValueError('eps was not defined, and qp.environ["SAMPLE_SIZE"] was not set')
-        eps = 1. / (2. * sample_size)
+        else:
+            eps = 1. / (2. * sample_size)
     return eps
 
 
@@ -246,8 +217,7 @@ CLASSIFICATION_ERROR_NAMES = {func.__name__ for func in CLASSIFICATION_ERROR}
 QUANTIFICATION_ERROR_NAMES = {func.__name__ for func in QUANTIFICATION_ERROR}
 QUANTIFICATION_ERROR_SINGLE_NAMES = {func.__name__ for func in QUANTIFICATION_ERROR_SINGLE}
 QUANTIFICATION_ERROR_SMOOTH_NAMES = {func.__name__ for func in QUANTIFICATION_ERROR_SMOOTH}
-ERROR_NAMES = \
+ERROR_NAMES = CLASSIFICATION_ERROR_NAMES | QUANTIFICATION_ERROR_NAMES | QUANTIFICATION_ERROR_SINGLE_NAMES
-    CLASSIFICATION_ERROR_NAMES | QUANTIFICATION_ERROR_NAMES | QUANTIFICATION_ERROR_SINGLE_NAMES
 
 f1_error = f1e
 acc_error = acce
@@ -255,3 +225,4 @@ mean_absolute_error = mae
 absolute_error = ae
 mean_relative_absolute_error = mrae
 relative_absolute_error = rae
+

quapy/method/aggregative.py

@@ -444,28 +444,24 @@ class EMQ(AggregativeProbabilisticQuantifier):
 
     def __init__(self, classifier: BaseEstimator, exact_train_prev=True, recalib=None):
         self.classifier = classifier
-        self.non_calibrated = classifier
         self.exact_train_prev = exact_train_prev
         self.recalib = recalib
 
     def fit(self, data: LabelledCollection, fit_classifier=True):
         if self.recalib is not None:
             if self.recalib == 'nbvs':
-                self.classifier = NBVSCalibration(self.non_calibrated)
+                self.classifier = NBVSCalibration(self.classifier)
             elif self.recalib == 'bcts':
-                self.classifier = BCTSCalibration(self.non_calibrated)
+                self.classifier = BCTSCalibration(self.classifier)
             elif self.recalib == 'ts':
-                self.classifier = TSCalibration(self.non_calibrated)
+                self.classifier = TSCalibration(self.classifier)
             elif self.recalib == 'vs':
-                self.classifier = VSCalibration(self.non_calibrated)
+                self.classifier = VSCalibration(self.classifier)
             elif self.recalib == 'platt':
                 self.classifier = CalibratedClassifierCV(self.classifier, ensemble=False)
             else:
                 raise ValueError('invalid param argument for recalibration method; available ones are '
                                  '"nbvs", "bcts", "ts", and "vs".')
-            self.recalib = None
-        else:
-            self.classifier = self.non_calibrated
         self.classifier, _ = _training_helper(self.classifier, data, fit_classifier, ensure_probabilistic=True)
         if self.exact_train_prev:
             self.train_prevalence = F.prevalence_from_labels(data.labels, self.classes_)
@@ -770,9 +766,7 @@ class DistributionMatching(AggregativeProbabilisticQuantifier):
         """
         Trains the classifier (if requested) and generates the validation distributions out of the training data.
         The validation distributions have shape `(n, ch, nbins)`, with `n` the number of classes, `ch` the number of
-        channels (a channel is a description, in form of a histogram, of a specific class -- there are as many channels
+        channels, and `nbins` the number of bins. In particular, let `V` be the validation distributions; `di=V[i]`
-        as classes, although in the binary case one can use only one channel, since the other one is constrained),
-        and `nbins` the number of bins. In particular, let `V` be the validation distributions; `di=V[i]`
         are the distributions obtained from training data labelled with class `i`; `dij = di[j]` is the discrete
         distribution of posterior probabilities `P(Y=j|X=x)` for training data labelled with class `i`, and `dij[k]`
         is the fraction of instances with a value in the `k`-th bin.
@@ -821,7 +815,7 @@ class DistributionMatching(AggregativeProbabilisticQuantifier):
         uniform_distribution = np.full(fill_value=1 / n_classes, shape=(n_classes,))
 
         # solutions are bounded to those contained in the unit-simplex
-        bounds = tuple((0, 1) for _ in range(n_classes))  # values in [0,1]
+        bounds = tuple((0, 1) for x in range(n_classes))  # values in [0,1]
         constraints = ({'type': 'eq', 'fun': lambda x: 1 - sum(x)})  # values summing up to 1
         r = optimize.minimize(match, x0=uniform_distribution, method='SLSQP', bounds=bounds, constraints=constraints)
         return r.x

quapy/method/neural.py

@@ -9,7 +9,6 @@ from torch.nn.functional import relu
 from quapy.protocol import UPP
 from quapy.method.aggregative import *
 from quapy.util import EarlyStop
-from tqdm import tqdm
 
 
 class QuaNetTrainer(BaseQuantifier):
@@ -29,7 +28,7 @@ class QuaNetTrainer(BaseQuantifier):
     >>>
     >>> # load the kindle dataset as text, and convert words to numerical indexes
     >>> dataset = qp.datasets.fetch_reviews('kindle', pickle=True)
-    >>> qp.domains.preprocessing.index(dataset, min_df=5, inplace=True)
+    >>> qp.data.preprocessing.index(dataset, min_df=5, inplace=True)
     >>>
     >>> # the text classifier is a CNN trained by NeuralClassifierTrainer
     >>> cnn = CNNnet(dataset.vocabulary_size, dataset.n_classes)
@@ -264,19 +263,15 @@ class QuaNetTrainer(BaseQuantifier):
                                      f'patience={early_stop.patience}/{early_stop.PATIENCE_LIMIT}')
 
     def get_params(self, deep=True):
-        classifier_params = self.classifier.get_params()
+        return {**self.classifier.get_params(), **self.quanet_params}
-        classifier_params = {'classifier__'+k:v for k,v in classifier_params.items()}
-        return {**classifier_params, **self.quanet_params}
 
     def set_params(self, **parameters):
         learner_params = {}
         for key, val in parameters.items():
             if key in self.quanet_params:
                 self.quanet_params[key] = val
-            elif key.startswith('classifier__'):
-                learner_params[key.replace('classifier__', '')] = val
             else:
-                raise ValueError('unknown parameter ', key)
+                learner_params[key] = val
         self.classifier.set_params(**learner_params)
 
     def __check_params_colision(self, quanet_params, learner_params):

quapy/model_selection.py

@@ -56,7 +56,7 @@ class GridSearchQ(BaseQuantifier):
 
     def _sout(self, msg):
         if self.verbose:
-            print(f'[{self.__class__.__name__}:{self.model.__class__.__name__}]: {msg}')
+            print(f'[{self.__class__.__name__}]: {msg}')
 
     def __check_error(self, error):
         if error in qp.error.QUANTIFICATION_ERROR:

quapy/plot.py

@@ -383,9 +383,6 @@ def brokenbar_supremacy_by_drift(method_names, true_prevs, estim_prevs, tr_prevs
     # x_error function) and 'y' is the estim-test shift (computed as according to y_error)
     data = _join_data_by_drift(method_names, true_prevs, estim_prevs, tr_prevs, x_error, y_error, method_order)
 
-    if method_order is None:
-        method_order = method_names
-
     if binning == 'isomerous':
         # take bins containing the same amount of examples
         tr_test_drifts = np.concatenate([data[m]['x'] for m in method_order])

setup.py

@@ -89,6 +89,8 @@ setup(
         'License :: OSI Approved :: BSD License',
 
         'Programming Language :: Python :: 3',
+        'Programming Language :: Python :: 3.6',
+        'Programming Language :: Python :: 3.7',
         'Programming Language :: Python :: 3.8',
         'Programming Language :: Python :: 3.9',
         'Programming Language :: Python :: 3 :: Only',
@@ -111,7 +113,7 @@ setup(
     #
     packages=find_packages(include=['quapy', 'quapy.*']),  # Required
 
-    python_requires='>=3.8, <4',
+    python_requires='>=3.6, <4',
 
     install_requires=['scikit-learn', 'pandas', 'tqdm', 'matplotlib', 'joblib', 'xlrd', 'abstention'],