doing tests with quantification stumps

NewMethods/class_weight_model.py

@@ -28,7 +28,8 @@ Possible extensions:
     - better investigate the "iterative" nature of the method.
     - better investigate the implications with other learners. E.g., using EMQ as a prompt, or using EMQ in the second
         stage (test).
-    - test with SVM
+    - test with SVM (not working well... and problematic due to the fact that svms need to be calibrated)
+    - test in multiclass scenarios
 """
 
 class ClassWeightPCC(BaseQuantifier):

NewMethods/reviews_experiments.py

@@ -42,7 +42,7 @@ def quantification_models():
     # yield 'pcc.opt', PCC(newLR()), lr_params
     # yield 'pacc.opt', PACC(newLR()), lr_params
     # yield 'wpacc.opt', ClassWeightPCC(), lr_params
-    yield 'ds', QuantificationStumpRegressor(SAMPLE_SIZE), None
+    yield 'ds', QuantificationStumpRegressor(SAMPLE_SIZE, 21, 10), None
     # yield 'ds.opt', QuantificationStumpRegressor(SAMPLE_SIZE), {'C': __C_range}
     # yield 'MAX', MAX(newLR()), lr_params
     # yield 'MS', MS(newLR()), lr_params
@@ -130,8 +130,8 @@ def run(experiment):
         model,
         test=data.test,
         sample_size=SAMPLE_SIZE,
-        n_prevpoints=21,
-        n_repetitions=1,
+        n_prevpoints=21,  # 21
+        n_repetitions=10,  # 100
         n_jobs=-1 if isinstance(model, qp.method.meta.Ensemble) else 1,
         verbose=True
     )

NewMethods/reviews_tables.py

@@ -5,7 +5,7 @@ import sys, os
 import pickle
 import argparse
 from common import *
-from tc_experiments import *
+from reviews_experiments import *
 from tabular import Table
 import itertools