# HG changeset patch
# User youssouf
# Date 1267713803 18000
# Node ID e12702b88a2d6fb360297b0e368a6048f86e451a
# Parent  70a9df1cd20e7232d524c9e921973fc4f98ff58c
removed one bug: the complexity value was not sent to the dA class

diff -r 70a9df1cd20e -r e12702b88a2d deep/autoencoder/DA_training.py
--- a/deep/autoencoder/DA_training.py	Tue Mar 02 09:52:27 2010 -0500
+++ b/deep/autoencoder/DA_training.py	Thu Mar 04 09:43:23 2010 -0500
@@ -93,7 +93,12 @@
         theano_rng = RandomStreams()
         # create a numpy random generator
         numpy_rng = numpy.random.RandomState()
-        
+		
+        # print the parameter of the DA
+        if True :
+            print 'input size = %d' %n_visible
+            print 'hidden size = %d' %n_hidden
+            print 'complexity = %2.2f' %complexity
          
         # initial values for weights and biases
         # note : W' was written as `W_prime` and b' as `b_prime`
@@ -250,7 +255,7 @@
 
     # construct the denoising autoencoder class
     n_ins = 32*32
-    encoder = dA(n_ins, n_code_layer, input = x.reshape((batch_size,n_ins)))
+    encoder = dA(n_ins, n_code_layer, complexity, input = x.reshape((batch_size,n_ins)))
 
     # Train autoencoder
     
@@ -363,7 +368,7 @@
                               test_score))
 
         if patience <= iter :
-                print('iter (%i) is superior than patience(%i). break', iter, patience)
+                print('iter (%i) is superior than patience(%i). break', (iter, patience))
                 break
 
         
@@ -451,7 +456,7 @@
 
     # construct the denoising autoencoder class
     n_ins = 28*28
-    encoder = dA(n_ins, n_code_layer, input = x.reshape((batch_size,n_ins)))
+    encoder = dA(n_ins, n_code_layer, complexity, input = x.reshape((batch_size,n_ins)))
 
     # Train autoencoder