How can I get hidden layer representation of the given data #169

Activity

1. tingjianlau commented on Oct 12, 2016

   

   tingjianlau

   on Oct 12, 2016

   Last edited by tingjianlau

   More actions

   In DBN.py, I add the extract_features() to get the output of the rbms.

   def extract_features(self, aim_set_x, batch_size):
           n_aim_batches = aim_set_x.get_value(borrow=True).shape[0]
           n_aim_batches //= batch_size
           index = T.lscalar('index')  # index to a [mini]batch
   
            # extract the feature
           extract_features_i = theano.function(
               [index],
               self.sigmoid_layers[-1].output,
               givens={
                   self.x: aim_set_x[
                       index * batch_size: (index + 1) * batch_size
                   ]
               }
           )      
           return [extract_features_i(i) for i in range(n_aim_batches)]