lorenzo.volpi
/
QuaPy
forked from moreo/QuaPy
1
0
Fork 0
Code Pull Requests Activity

trying to understand the poor performance of quantifiers with rel sampling

This commit is contained in:
Alejandro Moreo Fernandez 2022-02-03 11:00:25 +01:00
parent 6ea627449c
commit 1b1f41dc28
3 changed files with 8 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
eDiscovery/functions.py
View File

@ -49,8 +49,9 @@ def NewQuantifier(quantifiername, classifiername):
def newQ(): def newQ():
# return PACC(NewClassifier(classifiername), val_split=0.4) # return PACC(NewClassifier(classifiername), val_split=0.4)
# return CC(CalibratedClassifierCV(NewClassifier(classifiername))) # return CC(CalibratedClassifierCV(NewClassifier(classifiername)))
return ClassWeightPCC() # return ClassWeightPCC()
return RegionProbAdjustmentGlobal(newQ, k=10, clustering='kmeans') return CC(NewClassifier(classifiername))
return RegionProbAdjustmentGlobal(newQ, k=20, clustering='kmeans')
raise ValueError('unknown quantifier', quantifiername) raise ValueError('unknown quantifier', quantifiername)

6
eDiscovery/method.py
View File

@ -203,7 +203,7 @@ class RegionProbAdjustmentGlobal(BaseQuantifier):
# g = self._get_regions(data.instances) # g = self._get_regions(data.instances)
X, y = data.Xy X, y = data.Xy
self.g_quantifiers = {} self.g_quantifiers = {}
trivial=0 trivial, trivial_data = 0, 0
for gi in np.unique(g): for gi in np.unique(g):
qi_data = LabelledCollection(X[g==gi], y[g==gi], classes_=data.classes_) qi_data = LabelledCollection(X[g==gi], y[g==gi], classes_=data.classes_)
if qi_data.counts()[1] <= 1: if qi_data.counts()[1] <= 1:
@ -213,12 +213,14 @@ class RegionProbAdjustmentGlobal(BaseQuantifier):
# if qi_data.prevalence()[0] == 1: # all negatives # if qi_data.prevalence()[0] == 1: # all negatives
self.g_quantifiers[gi] = TrivialRejectorQuantifier() self.g_quantifiers[gi] = TrivialRejectorQuantifier()
trivial+=1 trivial+=1
trivial_data += len(qi_data)
elif qi_data.counts()[0] <= 1: # (almost) all positives elif qi_data.counts()[0] <= 1: # (almost) all positives
self.g_quantifiers[gi] = TrivialAcceptorQuantifier() self.g_quantifiers[gi] = TrivialAcceptorQuantifier()
trivial += 1 trivial += 1
trivial_data += len(qi_data)
else: else:
self.g_quantifiers[gi] = self.quantifier_fn().fit(qi_data) self.g_quantifiers[gi] = self.quantifier_fn().fit(qi_data)
print(f'trivials={trivial}') print(f'trivials={trivial} amounting to {trivial_data*100.0/len(data):.2f}% of the data')
return self return self

2
eDiscovery/run.sh
View File

@ -7,7 +7,7 @@ k=100
initsize=500 initsize=500
initprev=-1 initprev=-1
seed=1 seed=1
Q=GRPACC Q=URBQ
CLS=lr CLS=lr
sampling=relevance_sampling sampling=relevance_sampling