verbose flag

lkugler · Aug 7, 2013 · 2e32d43 · 2e32d43
1 parent a09e4d9
commit 2e32d43
Show file tree

Hide file tree

Showing 4 changed files with 26 additions and 9 deletions.
diff --git a/pegasos/__init__.py b/pegasos/__init__.py
@@ -22,24 +22,28 @@ class PegasosSVMClassifier(SVMPegasosBase):
     def __init__(self,
                  iterations=constants.DFLT_ITERATIONS,
                  lambda_reg=constants.DFLT_LAMBDA_REG,
-                 loop_type=constants.LOOP_STOCHASTIC):
+                 loop_type=constants.LOOP_STOCHASTIC,
+                 verbose=0):
 
         super(SVMPegasosBase, self).__init__(
                 iterations,
                 lambda_reg,
                 constants.LEARNER_PEGASOS_SVM,
-                loop_type)
+                loop_type,
+                verbose)
 
 
 class PegasosLogisticRegression(LogisticPegasosBase):
     def __init__(self,
                  iterations=constants.DFLT_ITERATIONS,
                  lambda_reg=constants.DFLT_LAMBDA_REG,
-                 loop_type=constants.LOOP_STOCHASTIC):
+                 loop_type=constants.LOOP_STOCHASTIC,
+                 verbose=0):
 
         super(PegasosLogisticRegression, self).__init__(
                 iterations,
                 lambda_reg,
                 constants.LEARNER_PEGASOS_LOGREG,
-                loop_type)
+                loop_type,
+                verbose)
 
diff --git a/pegasos/base.py b/pegasos/base.py
@@ -35,13 +35,16 @@ def __init__(self,
                  iterations,
                  lambda_reg,
                  learner_type,
-                 loop_type):
+                 loop_type,
+                 verbose):
 
         self.iterations = iterations
         self.lambda_reg = lambda_reg
         self.loop_type = loop_type
         self.learner_type = learner_type
 
+        self.verbose = verbose
+
         self.weight_vector = None
 
     def fit(self, X, y):

diff --git a/pegasos/models.py b/pegasos/models.py
@@ -30,7 +30,8 @@ def __init__(self,
                  iterations,
                  lambda_reg,
                  learner_type,
-                 loop_type):
+                 loop_type,
+                 verbose):
 
         if learner_type != constants.LEARNER_PEGASOS_SVM:
             raise ValueError('%s only supports SVM learners' % self.__class__.__name__)
@@ -39,7 +40,8 @@ def __init__(self,
                 iterations,
                 lambda_reg,
                 learner_type,
-                loop_type)
+                loop_type,
+                verbose)
 
 
 class LogisticPegasosBase(PegasosBase):
@@ -50,7 +52,8 @@ def __init__(self,
                  iterations,
                  lambda_reg,
                  learner_type,
-                 loop_type):
+                 loop_type,
+                 verbose):
 
         if learner_type != constants.LEARNER_PEGASOS_LOGREG:
             raise ValueError('%s only supports logistic learners' % self.__class__.__name__)
@@ -59,7 +62,8 @@ def __init__(self,
                 iterations,
                 lambda_reg,
                 learner_type,
-                loop_type)
+                loop_type,
+                verbose)
 
     def predict_proba(self, X):
         if not self.weight_vector:

diff --git a/pegasos/pegasos.py b/pegasos/pegasos.py
@@ -69,6 +69,9 @@ def train_stochastic(model, X, y):
         else:
             raise ValueError('%s: unknown learner type' % model.loop_type)
 
+        if model.verbose > 1 or (model.verbose == 1 and iteration % 1000 == 0):
+            print 'train_stochastic: i=%d' % iteration
+
 def train_stochastic_balanced(model, X, y):
     """
     At each training step we sample a negative and positive
@@ -99,6 +102,9 @@ def train_stochastic_balanced(model, X, y):
         else:
             raise ValueError('%s: unknown learner type' % model.loop_type)
 
+        if model.verbose > 1 or (model.verbose == 1 and iteration % 1000 == 0):
+            print 'train_stochastic_balanced: i=%d' % iteration
+
 def predict(model, X):
     if sparse.issparse(X):
         return np.array((model.weight_vector.weights * X.T).todense()).reshape(-1)