cvxopt_unforced

multiview_platform/MonoMultiViewClassifiers/Monoview/Additions/CQBoostUtils.py

@@ -186,7 +186,7 @@ class ColumnGenerationClassifier(BaseEstimator, ClassifierMixin, BaseBoost):
         if previous_w is not None:
             qp.initial_values = np.append(previous_w, [0])
 
-        # try:
+        try:
             solver_result = qp.solve(abstol=1e-10, reltol=1e-10, feastol=1e-10, return_all_information=True)
             w = np.asarray(np.array(solver_result['x']).T[0])[:n_hypotheses]
 
@@ -200,17 +200,17 @@ class ColumnGenerationClassifier(BaseEstimator, ClassifierMixin, BaseBoost):
                 self.dual_constraint_rhs = dual_variables[-1]
                 # logging.info('Updating dual constraint rhs: {}'.format(self.dual_constraint_rhs))
 
-        # except:
-        #     logging.warning('QP Solving failed at iteration {}.'.format(n_hypotheses))
-        #     if previous_w is not None:
-        #         w = np.append(previous_w, [0])
-        #     else:
-        #         w = np.array([1.0 / n_hypotheses] * n_hypotheses)
-        #
-        #     if previous_alpha is not None:
-        #         alpha = previous_alpha
-        #     else:
-        #         alpha = self._initialize_alphas(n_examples)
+        except:
+            logging.warning('QP Solving failed at iteration {}.'.format(n_hypotheses))
+            if previous_w is not None:
+                w = np.append(previous_w, [0])
+            else:
+                w = np.array([1.0 / n_hypotheses] * n_hypotheses)
+
+            if previous_alpha is not None:
+                alpha = previous_alpha
+            else:
+                alpha = self._initialize_alphas(n_examples)
 
         return w, alpha