merging

Author: Razvan Pascanu

code/DBN.py

@@ -2,7 +2,7 @@
 """
 import os
 
-import numpy, time, cPickle, gzip 
+import numpy, time, cPickle, gzip, os, sys
 
 import theano
 import theano.tensor as T
@@ -141,6 +141,7 @@ def pretraining_functions(self, train_set_x, batch_size):
         # index to a [mini]batch
         index            = T.lscalar('index')   # index to a minibatch
         learning_rate    = T.scalar('lr')    # learning rate to use
+        k                = T.lscalar('k')
 
         # number of batches
         n_batches = train_set_x.value.shape[0] / batch_size
@@ -154,11 +155,12 @@ def pretraining_functions(self, train_set_x, batch_size):
 
             # get the cost and the updates list
             # TODO: change cost function to reconstruction error
-            cost,updates = rbm.cd(learning_rate, persistent=None)
+            cost,updates = rbm.get_cost_updates(learning_rate, persistent=None, k =k)
 
             # compile the theano function    
             fn = theano.function(inputs = [index, 
-                              theano.Param(learning_rate, default = 0.1)], 
+                              theano.Param(learning_rate, default = 0.1),
+                              theano.Param(k, default = 1)],
                     outputs = cost, 
                     updates = updates,
                     givens  = {self.x :train_set_x[batch_begin:batch_end]})
@@ -229,13 +231,10 @@ def test_score():
         return train_fn, valid_score, test_score
 
 
-
-
-
-
 def test_DBN( finetune_lr = 0.1, pretraining_epochs = 10, \
-              pretrain_lr = 0.1, training_epochs = 1000, \
-              dataset='mnist.pkl.gz'):
+              pretrain_lr = 0.1, k = 1, training_epochs = 1000, \
+              dataset='../data/mnist.pkl.gz', batch_size = 1,
+              output_folder = 'DBN_plots'):
     """
     Demonstrates how to train and test a Deep Belief Network.
 
@@ -253,18 +252,13 @@ def test_DBN( finetune_lr = 0.1, pretraining_epochs = 10, \
     :param dataset: path the the pickled dataset
     """
 
-    print 'finetune_lr = ', finetune_lr
-    print 'pretrain_lr = ', pretrain_lr
 
     datasets = load_data(dataset)
 
     train_set_x, train_set_y = datasets[0]
     valid_set_x, valid_set_y = datasets[1]
     test_set_x , test_set_y  = datasets[2]
 
-
-    batch_size = 20    # size of the minibatch
-
     # compute number of minibatches for training, validation and testing
     n_train_batches = train_set_x.value.shape[0] / batch_size
 
@@ -295,7 +289,7 @@ def test_DBN( finetune_lr = 0.1, pretraining_epochs = 10, \
             c = []
             for batch_index in xrange(n_train_batches):
                 c.append(pretraining_fns[i](index = batch_index, 
-                         lr = pretrain_lr ) )
+                         lr = pretrain_lr, k = k ) )
             print 'Pre-training layer %i, epoch %d, cost '%(i,epoch),numpy.mean(c)
 
     end_time = time.clock()

code/SdA.py

@@ -281,7 +281,7 @@ def test_score():
 
 def test_SdA( finetune_lr = 0.1, pretraining_epochs = 15, \
               pretrain_lr = 0.05, training_epochs = 1000, \
-              dataset='mnist.pkl.gz'):
+              dataset='../data/mnist.pkl.gz', batch_size = 1):
     """
     Demonstrates how to train and test a stochastic denoising autoencoder.
 
@@ -312,8 +312,6 @@ def test_SdA( finetune_lr = 0.1, pretraining_epochs = 15, \
     test_set_x , test_set_y  = datasets[2]
 
 
-    batch_size = 20    # size of the minibatch
-
     # compute number of minibatches for training, validation and testing
     n_train_batches = train_set_x.value.shape[0] / batch_size
 

code/convolutional_mlp.py

@@ -96,7 +96,8 @@ def __init__(self, rng, input, filter_shape, image_shape, poolsize=(2,2)):
 
 
 
-def evaluate_lenet5(learning_rate=0.1, n_epochs=200, dataset='mnist.pkl.gz', nkerns=[20,50]):
+def evaluate_lenet5(learning_rate=0.1, n_epochs=200, dataset='../data/mnist.pkl.gz',
+        nkerns=[20,50], batch_size = 500):
     """ Demonstrates lenet on MNIST dataset
 
     :type learning_rate: float
@@ -122,8 +123,6 @@ def evaluate_lenet5(learning_rate=0.1, n_epochs=200, dataset='mnist.pkl.gz', nke
     test_set_x , test_set_y  = datasets[2]
 
 
-    batch_size = 500    # size of the minibatch
-
     # compute number of minibatches for training, validation and testing
     n_train_batches = train_set_x.value.shape[0] / batch_size
     n_valid_batches = valid_set_x.value.shape[0] / batch_size

code/dA.py

@@ -129,8 +129,8 @@ def __init__(self, numpy_rng, theano_rng = None, input = None, n_visible= 784, n
             # converted using asarray to dtype
             # theano.config.floatX so that the code is runable on GPU
             initial_W = numpy.asarray( numpy_rng.uniform( 
-                      low  = -numpy.sqrt(6./(n_hidden+n_visible)), 
-                      high = numpy.sqrt(6./(n_hidden+n_visible)), 
+                      low  = -4*numpy.sqrt(6./(n_hidden+n_visible)), 
+                      high =  4*numpy.sqrt(6./(n_hidden+n_visible)), 
                       size = (n_visible, n_hidden)), dtype = theano.config.floatX)
             W = theano.shared(value = initial_W, name ='W')
 
@@ -219,7 +219,8 @@ def get_cost_updates(self, corruption_level, learning_rate):
 
 
 
-def test_dA( learning_rate = 0.1, training_epochs = 15, dataset ='mnist.pkl.gz' ):
+def test_dA( learning_rate = 0.1, training_epochs = 15, dataset ='../data/mnist.pkl.gz',
+        batch_size = 20, output_folder = 'dA_plots' ):
 
     """
     This demo is tested on MNIST
@@ -237,15 +238,17 @@ def test_dA( learning_rate = 0.1, training_epochs = 15, dataset ='mnist.pkl.gz'
     datasets = load_data(dataset)
     train_set_x, train_set_y = datasets[0]
 
-    batch_size = 20   # size of the minibatch
-
     # compute number of minibatches for training, validation and testing
     n_train_batches = train_set_x.value.shape[0] / batch_size
 
     # allocate symbolic variables for the data
     index = T.lscalar()    # index to a [mini]batch 
     x     = T.matrix('x')  # the data is presented as rasterized images
 
+    
+    if not os.path.isdir(output_folder):
+        os.makedirs(output_folder)
+    os.chdir(output_folder)
     ####################################
     # BUILDING THE MODEL NO CORRUPTION #
     ####################################
@@ -331,6 +334,7 @@ def test_dA( learning_rate = 0.1, training_epochs = 15, dataset ='mnist.pkl.gz'
                  tile_spacing=(1,1)))
     image.save('filters_corruption_30.png') 
 
+    os.chdir('../')
 
 
 if __name__ == '__main__':