def evaluate_lenet5(datasets, learning_seed=0.01, n_epochs=500, batch_size=250, save_folder='./cache', channel_count=1): """ Evaluate a convnet for three dimensional image inputs. :type learning_seed: float :param learning_seed: learning rate used (factor for the stochastic gradient) during initialization. :type n_epochs: int :param n_epochs: maximal number of epochs to run the optimizer :type dataset: string :param dataset: path to the dataset used for training /testing (MNIST here) :type batch_size: integer :param batch_size: size for batched testing :type channel_count: integer :param channel_count: number of channels per image """ rng = numpy.random.RandomState(23455) train_set_x, train_set_y = datasets[0] valid_set_x, valid_set_y = datasets[1] test_set_x, test_set_y = datasets[2] # compute number of minibatches for training, validation and testing n_train_batches = train_set_x.get_value(borrow=True).shape[0] n_valid_batches = valid_set_x.get_value(borrow=True).shape[0] n_test_batches = test_set_x.get_value(borrow=True).shape[0] n_train_batches /= batch_size n_valid_batches /= batch_size n_test_batches /= batch_size # allocate symbolic variables for the data index = T.lscalar() # index to a [mini]batch new_rate = T.lscalar() # The learning rate. # start-snippet-1 r = T.dscalar('r') # the learning rate as a variable. x = T.matrix('x') # the data is presented as rasterized images y = T.ivector('y') # the labels are presented as 1D vector of # [int] labels ###################### # BUILD ACTUAL MODEL # ###################### print '... building the model' # Reshape matrix of rasterized images of shape # (batch_size, channel_count, 32 * 32) # to a 4D tensor, compatible with our LeNetConvPoolLayer # (32, 32) is the size of CIFAR images. layer0_input = x.reshape((batch_size, channel_count, 32, 32)) # Construct the first convolutional pooling layer: # filtering reduces the image size to (32+2-3+1 , 32+2-3+1) = (32, 32) # maxpooling reduces this further to (32/2, 32/2) = (16, 16) # 4D output tensor is thus of shape (batch_size, nkerns[0], 16, 16) layer0 = LeNetConvPoolLayer( rng, input=layer0_input, image_shape=(batch_size, channel_count, 32, 32), filter_shape=(128, channel_count, 3, 3), poolsize=(1, 1) ) layer1 = LeNetConvPoolLayer( rng, input=layer0.output, image_shape=(batch_size, 128, 32, 32), filter_shape=(128, 128, 3, 3), poolsize=(2, 2) ) # Construct the second convolutional pooling layer # filtering reduces the image size to (16+2-3+1, 16+2-3+1) = (16, 16) # maxpooling reduces this further to (16/2, 16/2) = (8, 8) # 4D output tensor is thus of shape (batch_size, nkerns[1], 8, 8) layer2 = LeNetConvPoolLayer( rng, input=layer1.output, image_shape=(batch_size, 128, 16, 16), filter_shape=(256, 128, 3, 3), poolsize=(1, 1) ) layer3 = LeNetConvPoolLayer( rng, input=layer2.output, image_shape=(batch_size, 256, 16, 16), filter_shape=(256, 256, 3, 3), poolsize=(1, 1) ) layer4 = LeNetConvPoolLayer( rng, input=layer3.output, image_shape=(batch_size, 256, 16, 16), filter_shape=(256, 256, 3, 3), poolsize=(1, 1) ) layer5 = LeNetConvPoolLayer( rng, input=layer4.output, image_shape=(batch_size, 256, 16, 16), filter_shape=(256, 256, 3, 3), poolsize=(2, 2) ) # Construct the third convolutional pooling layer # filtering reduces the image size to (8+2-3+1, 8+2-3+1) = (8, 8) # No maxpooling (aka maxpooling (1, 1)) # 4D output tensor is thus of shape (batch_size, nkerns[1], 8, 8) layer6 = LeNetConvPoolLayer( rng, input=layer5.output, image_shape=(batch_size, 256, 8, 8), filter_shape=(512, 256, 3, 3), poolsize=(1, 1) ) layer7 = LeNetConvPoolLayer( rng, input=layer6.output, image_shape=(batch_size, 512, 8, 8), filter_shape=(512, 512, 3, 3), poolsize=(1, 1) ) layer8 = LeNetConvPoolLayer( rng, input=layer7.output, image_shape=(batch_size, 512, 8, 8), filter_shape=(512, 512, 3, 3), poolsize=(1, 1) ) layer9 = LeNetConvPoolLayer( rng, input=layer8.output, image_shape=(batch_size, 512, 8, 8), filter_shape=(512, 512, 3, 3), poolsize=(2, 2) ) # Construct the third convolutional pooling layer # filtering reduces the image size to (8+2-3+1, 8+2-3+1) = (8, 8) # maxpooling reduces this further to (8/2, 8/2) = (4, 4) # 4D output tensor is thus of shape (batch_size, nkerns[1], 4, 4) layer10 = LeNetConvPoolLayer( rng, input=layer9.output, image_shape=(batch_size, 512, 4, 4), filter_shape=(512, 512, 3, 3), poolsize=(1, 1) ) layer11 = LeNetConvPoolLayer( rng, input=layer10.output, image_shape=(batch_size, 512, 4, 4), filter_shape=(512, 512, 3, 3), poolsize=(1, 1) ) layer12 = LeNetConvPoolLayer( rng, input=layer11.output, image_shape=(batch_size, 512, 4, 4), filter_shape=(512, 512, 3, 3), poolsize=(1, 1) ) layer13 = LeNetConvPoolLayer( rng, input=layer12.output, image_shape=(batch_size, 512, 4, 4), filter_shape=(512, 512, 3, 3), poolsize=(1, 1) ) # the HiddenLayer being fully-connected, it operates on 2D matrices of # shape (batch_size, num_pixels) (i.e matrix of rasterized images). # This will generate a matrix of shape (500, 512 * 4 * 4) = (500, 8192) with the default values. layer14_input = layer13.output.flatten(2) # construct a fully-connected sigmoidal layer layer14 = HiddenLayer( rng, input=layer14_input, n_in=512 * 4 * 4, n_out=2048, activation=relu ) layer15 = HiddenLayer( rng, input=layer14.output, n_in=2048, n_out=1024, activation=relu ) # classify the values of the fully-connected sigmoidal layer # there are 10 labels in total. layer16 = HiddenLayer( rng, input=layer15.output, n_in=1024, n_out=10, activation=relu ) # the cost we minimize during training is the NLL of the model cost = layer16.negative_log_likelihood(y) # create a function to compute the mistakes that are made by the model test_model = theano.function( [index], layer16.errors(y), givens={ x: test_set_x[index * batch_size: (index + 1) * batch_size], y: test_set_y[index * batch_size: (index + 1) * batch_size] } ) validate_model = theano.function( [index], layer16.errors(y), givens={ x: valid_set_x[index * batch_size: (index + 1) * batch_size], y: valid_set_y[index * batch_size: (index + 1) * batch_size] } ) # create a list of all model parameters to be fit by gradient descent params = layer16.params + \ layer15.params + \ layer14.params + \ layer13.params + \ layer12.params + \ layer11.params + \ layer10.params + \ layer9.params + \ layer8.params + \ layer7.params + \ layer6.params + \ layer6.params + \ layer5.params + \ layer4.params + \ layer3.params + \ layer2.params + \ layer1.params + \ layer0.params # create a list of gradients for all model parameters grads = T.grad(cost, params) # train_model is a function that updates the model parameters by # SGD Since this model has many parameters, it would be tedious to # manually create an update rule for each model parameter. We thus # create the updates list by automatically looping over all # (params[i], grads[i]) pairs. updates = [ (param_i, param_i - r * grad_i) for param_i, grad_i in zip(params, grads) ] train_model = theano.function( [index, new_rate], cost, updates=updates, givens={ x: train_set_x[index * batch_size: (index + 1) * batch_size], y: train_set_y[index * batch_size: (index + 1) * batch_size], r: new_rate } ) # end-snippet-1 ############### # TRAIN MODEL # ############### print '... training' # early-stopping parameters patience = 10000 # look as this many examples regardless patience_increase = 2 # wait this much longer when a new best is # found improvement_threshold = 0.995 # a relative improvement of this much is # considered significant validation_frequency = min(n_train_batches, patience / 2) # go through this many # minibatche before checking the network # on the validation set; in this case we # check every epoch best_validation_loss = numpy.inf best_iter = 0 cur_learning_rate = learning_seed test_score = 0. start_time = timeit.default_timer() epoch = 0 done_looping = False while (epoch < n_epochs) and (not done_looping): epoch = epoch + 1 for minibatch_index in xrange(n_train_batches): iter = (epoch - 1) * n_train_batches + minibatch_index if iter % 100 == 0: print 'training @ iter = ', iter cost_ij = train_model(minibatch_index, cur_learning_rate) if (iter + 1) % validation_frequency == 0: # compute zero-one loss on validation set validation_losses = [validate_model(i) for i in xrange(n_valid_batches)] this_validation_loss = numpy.mean(validation_losses) print('epoch %i, minibatch %i/%i, validation error %f %%' % (epoch, minibatch_index + 1, n_train_batches, this_validation_loss * 100.)) # if we got the best validation score until now if this_validation_loss < best_validation_loss: #improve patience if loss improvement is good enough if this_validation_loss < best_validation_loss * \ improvement_threshold: patience = max(patience, iter * patience_increase) # save best validation score and iteration number best_validation_loss = this_validation_loss best_iter = iter # test it on the test set test_losses = [ test_model(i) for i in xrange(n_test_batches) ] test_score = numpy.mean(test_losses) print((' epoch %i, minibatch %i/%i, test error of ' 'best model %f %%') % (epoch, minibatch_index + 1, n_train_batches, test_score * 100.)) else: # Did not get a new best validation score. cur_learning_rate /= 10 if patience <= iter: done_looping = True break end_time = timeit.default_timer() print('Optimization complete.') print('Best validation score of %f %% obtained at iteration %i, ' 'with test performance %f %%' % (best_validation_loss * 100., best_iter + 1, test_score * 100.)) print >> sys.stderr, ('The code for file ' + os.path.split(__file__)[1] + ' ran for %.2fm' % ((end_time - start_time) / 60.))