--- a/deep/stacked_dae/v_sylvain/stacked_dae.py	Sun Apr 11 19:52:35 2010 -0400
+++ b/deep/stacked_dae/v_sylvain/stacked_dae.py	Sun Apr 11 19:52:44 2010 -0400
@@ -203,6 +203,7 @@
         self.x  = T.matrix('x')  # the data is presented as rasterized images
         self.y  = T.ivector('y') # the labels are presented as 1D vector of 
                                  # [int] labels
+        self.finetune_lr = T.fscalar('finetune_lr') #To get a dynamic finetune learning rate
 
         for i in xrange( self.n_layers ):
             # construct the sigmoidal layer
@@ -275,9 +276,9 @@
         # compute list of updates
         updates = {}
         for param,gparam in zip(self.params, gparams):
-            updates[param] = param - gparam*finetune_lr
+            updates[param] = param - gparam*self.finetune_lr
             
-        self.finetune = theano.function([self.x,self.y], cost, 
+        self.finetune = theano.function([self.x,self.y,self.finetune_lr], cost, 
                 updates = updates)#,
 
         # symbolic variable that points to the number of errors made on the
@@ -302,15 +303,16 @@
 
         #self.logistic_params+= self.logLayer2.params
         # construct a function that implements one step of finetunining
-
+        
+        self.logistic_params+=self.logLayer2.params
         # compute the cost, defined as the negative log likelihood 
         cost2 = self.logLayer2.negative_log_likelihood(self.y)
         # compute the gradients with respect to the model parameters
-        gparams2 = T.grad(cost2, self.logLayer2.params)
+        gparams2 = T.grad(cost2, self.logistic_params)
 
         # compute list of updates
         updates2 = {}
-        for param,gparam in zip(self.logLayer2.params, gparams2):
+        for param,gparam in zip(self.logistic_params, gparams2):
             updates2[param] = param - gparam*finetune_lr
    
         self.finetune2 = theano.function([self.x,self.y], cost2,