Mercurial > pylearn

changeset 501:4fb6f7320518

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
N-class logistic regression top-layer works
author Joseph Turian <turian@gmail.com>
date Tue, 28 Oct 2008 13:54:01 -0400
parents 3c60c2db0319
children 17945defd813
files algorithms/logistic_regression.py algorithms/tests/test_daa.py
diffstat 2 files changed, 26 insertions(+), 15 deletions(-) [+]
line wrap: on
line diff
--- a/algorithms/logistic_regression.py	Tue Oct 28 13:36:27 2008 -0400
+++ b/algorithms/logistic_regression.py	Tue Oct 28 13:54:01 2008 -0400
@@ -8,8 +8,11 @@
 import numpy as N
 
 class LogRegInstanceType(module.FancyModuleInstance):
-    def initialize(self, n_in, n_out=1, rng=N.random):
+    def initialize(self, n_in, n_out=1, rng=N.random, seed=None):
         #self.component is the LogisticRegressionTemplate instance that built this guy.
+        """
+        @todo: Remove seed. Used only to keep Stacker happy.
+        """
 
         self.w = N.zeros((n_in, n_out))
         self.b = N.zeros(n_out)
--- a/algorithms/tests/test_daa.py	Tue Oct 28 13:36:27 2008 -0400
+++ b/algorithms/tests/test_daa.py	Tue Oct 28 13:54:01 2008 -0400
@@ -23,21 +23,25 @@
     model.layers[2].noise_level = 0.3
 
     # Update the first hidden layer
-    model.local_update[0]([[0, 1, 0, 1]])
-    model.local_update[1]([[0, 1, 0, 1]])
-    model.local_update[2]([[0, 1, 0, 1]])
+    for l in range(3):
+        for i in range(10):
+            model.local_update[l]([[0, 1, 0, 1]])
+            model.local_update[l]([[1, 0, 1, 0]])
 
-    model.update([[0, 1, 0, 1]], [[0]])
+    for i in range(1):
+        model.update([[0, 1, 0, 1]], [[1]])
+        model.update([[1, 0, 1, 0]], [[0]])
     print model.classify([[0, 1, 0, 1]])
+    print model.classify([[1, 0, 1, 0]])
 
 
 def test_train_daa2(mode = theano.Mode('c|py', 'fast_run')):
 
     ndaa = 3
-    daa = models.Stacker([(models.SigmoidXEDenoisingAA, 'hidden')] * ndaa + [(pylearn.algorithms.logistic_regression.Module_Nclass, 'output')],
+    daa = models.Stacker([(models.SigmoidXEDenoisingAA, 'hidden')] * ndaa + [(pylearn.algorithms.logistic_regression.Module_Nclass, 'pred')],
                          regularize = False)
 
-    model = daa.make([4, 20, 20, 20, 1],
+    model = daa.make([4, 20, 20, 20, 10],
                      lr = 0.01,
                      mode = mode,
                      seed = 10)
@@ -46,14 +50,16 @@
     model.layers[1].noise_level = 0.3
     model.layers[2].noise_level = 0.3
 
-    # Update the first hidden layer
-    model.local_update[0]([[0, 1, 0, 1]])
-    model.local_update[1]([[0, 1, 0, 1]])
-    model.local_update[2]([[0, 1, 0, 1]])
+    for l in range(3):
+        for i in range(10):
+            model.local_update[l]([[0, 1, 0, 1]])
+            model.local_update[l]([[1, 0, 1, 0]])
 
-    model.update([[0, 1, 0, 1]], [0])
-    print model.classify([[0, 1, 0, 1]])
-
+    for i in range(1):
+        model.update([[0, 1, 0, 1]], [1])
+        model.update([[1, 0, 1, 0]], [0])
+    print model.apply([[0, 1, 0, 1]])
+    print model.apply([[1, 0, 1, 0]])
 
 
 
@@ -70,5 +76,7 @@
 ##    print 'time:',t2
 
 #    test_train_daa(theano.compile.Mode('c&py', 'merge'))
-    test_train_daa(theano.compile.Mode('c|py', 'merge'))
+#    test_train_daa(theano.compile.Mode('c|py', 'merge'))
+    test_train_daa(theano.compile.Mode('py', 'merge'))
+
     test_train_daa2(theano.compile.Mode('c|py', 'merge'))