Mercurial > ift6266
view code_tutoriel/rbm.py @ 352:cfb79f9fd1a4
Ajout d'une fonctionnalite pour pouvoir avoir un taux d'apprentissage decroissant dans le pretrain
author | SylvainPL <sylvain.pannetier.lebeuf@umontreal.ca> |
---|---|
date | Wed, 21 Apr 2010 14:50:59 -0400 |
parents | 4bc5eeec6394 |
children |
line wrap: on
line source
"""This tutorial introduces restricted boltzmann machines (RBM) using Theano. Boltzmann Machines (BMs) are a particular form of energy-based model which contain hidden variables. Restricted Boltzmann Machines further restrict BMs to those without visible-visible and hidden-hidden connections. """ import numpy, time, cPickle, gzip, PIL.Image import theano import theano.tensor as T import os from theano.tensor.shared_randomstreams import RandomStreams from utils import tile_raster_images from logistic_sgd import load_data class RBM(object): """Restricted Boltzmann Machine (RBM) """ def __init__(self, input=None, n_visible=784, n_hidden=500, \ W = None, hbias = None, vbias = None, numpy_rng = None, theano_rng = None): """ RBM constructor. Defines the parameters of the model along with basic operations for inferring hidden from visible (and vice-versa), as well as for performing CD updates. :param input: None for standalone RBMs or symbolic variable if RBM is part of a larger graph. :param n_visible: number of visible units :param n_hidden: number of hidden units :param W: None for standalone RBMs or symbolic variable pointing to a shared weight matrix in case RBM is part of a DBN network; in a DBN, the weights are shared between RBMs and layers of a MLP :param hbias: None for standalone RBMs or symbolic variable pointing to a shared hidden units bias vector in case RBM is part of a different network :param vbias: None for standalone RBMs or a symbolic variable pointing to a shared visible units bias """ self.n_visible = n_visible self.n_hidden = n_hidden if W is None : # W is initialized with `initial_W` which is uniformely sampled # from -6./sqrt(n_visible+n_hidden) and 6./sqrt(n_hidden+n_visible) # the output of uniform if converted using asarray to dtype # theano.config.floatX so that the code is runable on GPU initial_W = numpy.asarray( numpy.random.uniform( low = -numpy.sqrt(6./(n_hidden+n_visible)), high = numpy.sqrt(6./(n_hidden+n_visible)), size = (n_visible, n_hidden)), dtype = theano.config.floatX) # theano shared variables for weights and biases W = theano.shared(value = initial_W, name = 'W') if hbias is None : # create shared variable for hidden units bias hbias = theano.shared(value = numpy.zeros(n_hidden, dtype = theano.config.floatX), name='hbias') if vbias is None : # create shared variable for visible units bias vbias = theano.shared(value =numpy.zeros(n_visible, dtype = theano.config.floatX),name='vbias') if numpy_rng is None: # create a number generator numpy_rng = numpy.random.RandomState(1234) if theano_rng is None : theano_rng = RandomStreams(numpy_rng.randint(2**30)) # initialize input layer for standalone RBM or layer0 of DBN self.input = input if input else T.dmatrix('input') self.W = W self.hbias = hbias self.vbias = vbias self.theano_rng = theano_rng # **** WARNING: It is not a good idea to put things in this list # other than shared variables created in this function. self.params = [self.W, self.hbias, self.vbias] self.batch_size = self.input.shape[0] def free_energy(self, v_sample): ''' Function to compute the free energy ''' wx_b = T.dot(v_sample, self.W) + self.hbias vbias_term = T.sum(T.dot(v_sample, self.vbias)) hidden_term = T.sum(T.log(1+T.exp(wx_b))) return -hidden_term - vbias_term def sample_h_given_v(self, v0_sample): ''' This function infers state of hidden units given visible units ''' # compute the activation of the hidden units given a sample of the visibles h1_mean = T.nnet.sigmoid(T.dot(v0_sample, self.W) + self.hbias) # get a sample of the hiddens given their activation h1_sample = self.theano_rng.binomial(size = h1_mean.shape, n = 1, prob = h1_mean) return [h1_mean, h1_sample] def sample_v_given_h(self, h0_sample): ''' This function infers state of visible units given hidden units ''' # compute the activation of the visible given the hidden sample v1_mean = T.nnet.sigmoid(T.dot(h0_sample, self.W.T) + self.vbias) # get a sample of the visible given their activation v1_sample = self.theano_rng.binomial(size = v1_mean.shape,n = 1,prob = v1_mean) return [v1_mean, v1_sample] def gibbs_hvh(self, h0_sample): ''' This function implements one step of Gibbs sampling, starting from the hidden state''' v1_mean, v1_sample = self.sample_v_given_h(h0_sample) h1_mean, h1_sample = self.sample_h_given_v(v1_sample) return [v1_mean, v1_sample, h1_mean, h1_sample] def gibbs_vhv(self, v0_sample): ''' This function implements one step of Gibbs sampling, starting from the visible state''' h1_mean, h1_sample = self.sample_h_given_v(v0_sample) v1_mean, v1_sample = self.sample_v_given_h(h1_sample) return [h1_mean, h1_sample, v1_mean, v1_sample] def cd(self, lr = 0.1, persistent=None): """ This functions implements one step of CD-1 or PCD-1 :param lr: learning rate used to train the RBM :param persistent: None for CD. For PCD, shared variable containing old state of Gibbs chain. This must be a shared variable of size (batch size, number of hidden units). Returns the updates dictionary. The dictionary contains the update rules for weights and biases but also an update of the shared variable used to store the persistent chain, if one is used. """ # compute positive phase ph_mean, ph_sample = self.sample_h_given_v(self.input) # decide how to initialize persistent chain: # for CD, we use the newly generate hidden sample # for PCD, we initialize from the old state of the chain if persistent is None: chain_start = ph_sample else: chain_start = persistent # perform actual negative phase [nv_mean, nv_sample, nh_mean, nh_sample] = self.gibbs_hvh(chain_start) # determine gradients on RBM parameters g_vbias = T.sum( self.input - nv_mean, axis = 0)/self.batch_size g_hbias = T.sum( ph_mean - nh_mean, axis = 0)/self.batch_size g_W = T.dot(ph_mean.T, self.input )/ self.batch_size - \ T.dot(nh_mean.T, nv_mean )/ self.batch_size gparams = [g_W.T, g_hbias, g_vbias] # constructs the update dictionary updates = {} for gparam, param in zip(gparams, self.params): updates[param] = param + gparam * lr if persistent: # Note that this works only if persistent is a shared variable updates[persistent] = T.cast(nh_sample, dtype=theano.config.floatX) # pseudo-likelihood is a better proxy for PCD cost = self.get_pseudo_likelihood_cost(updates) else: # reconstruction cross-entropy is a better proxy for CD cost = self.get_reconstruction_cost(updates, nv_mean) return cost, updates def get_pseudo_likelihood_cost(self, updates): """Stochastic approximation to the pseudo-likelihood""" # index of bit i in expression p(x_i | x_{\i}) bit_i_idx = theano.shared(value=0, name = 'bit_i_idx') # binarize the input image by rounding to nearest integer xi = T.iround(self.input) # calculate free energy for the given bit configuration fe_xi = self.free_energy(xi) # flip bit x_i of matrix xi and preserve all other bits x_{\i} # Equivalent to xi[:,bit_i_idx] = 1-xi[:, bit_i_idx] # NB: slice(start,stop,step) is the python object used for # slicing, e.g. to index matrix x as follows: x[start:stop:step] xi_flip = T.setsubtensor(xi, 1-xi[:, bit_i_idx], idx_list=(slice(None,None,None),bit_i_idx)) # calculate free energy with bit flipped fe_xi_flip = self.free_energy(xi_flip) # equivalent to e^(-FE(x_i)) / (e^(-FE(x_i)) + e^(-FE(x_{\i}))) cost = self.n_visible * T.log(T.nnet.sigmoid(fe_xi_flip - fe_xi)) # increment bit_i_idx % number as part of updates updates[bit_i_idx] = (bit_i_idx + 1) % self.n_visible return cost def get_reconstruction_cost(self, updates, nv_mean): """Approximation to the reconstruction error""" cross_entropy = T.mean( T.sum(self.input*T.log(nv_mean) + (1 - self.input)*T.log(1-nv_mean), axis = 1)) return cross_entropy def test_rbm(learning_rate=0.1, training_epochs = 15, dataset='mnist.pkl.gz'): """ Demonstrate *** This is demonstrated on MNIST. :param learning_rate: learning rate used for training the RBM :param training_epochs: number of epochs used for training :param dataset: path the the pickled dataset """ datasets = load_data(dataset) train_set_x, train_set_y = datasets[0] 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 # 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 rng = numpy.random.RandomState(123) theano_rng = RandomStreams( rng.randint(2**30)) # initialize storage fot the persistent chain (state = hidden layer of chain) persistent_chain = theano.shared(numpy.zeros((batch_size, 500))) # construct the RBM class rbm = RBM( input = x, n_visible=28*28, \ n_hidden = 500,numpy_rng = rng, theano_rng = theano_rng) # get the cost and the gradient corresponding to one step of CD cost, updates = rbm.cd(lr=learning_rate, persistent=persistent_chain) ################################# # Training the RBM # ################################# dirname = 'lr=%.5f'%learning_rate os.makedirs(dirname) os.chdir(dirname) # it is ok for a theano function to have no output # the purpose of train_rbm is solely to update the RBM parameters train_rbm = theano.function([index], cost, updates = updates, givens = { x: train_set_x[index*batch_size:(index+1)*batch_size]}) plotting_time = 0. start_time = time.clock() # go through training epochs for epoch in xrange(training_epochs): # go through the training set mean_cost = [] for batch_index in xrange(n_train_batches): mean_cost += [train_rbm(batch_index)] print 'Training epoch %d, cost is '%epoch, numpy.mean(mean_cost) # Plot filters after each training epoch plotting_start = time.clock() # Construct image from the weight matrix image = PIL.Image.fromarray(tile_raster_images( X = rbm.W.value.T, img_shape = (28,28),tile_shape = (10,10), tile_spacing=(1,1))) image.save('filters_at_epoch_%i.png'%epoch) plotting_stop = time.clock() plotting_time += (plotting_stop - plotting_start) end_time = time.clock() pretraining_time = (end_time - start_time) - plotting_time print ('Training took %f minutes' %(pretraining_time/60.)) ################################# # Sampling from the RBM # ################################# # find out the number of test samples number_of_test_samples = test_set_x.value.shape[0] # pick random test examples, with which to initialize the persistent chain test_idx = rng.randint(number_of_test_samples-20) persistent_vis_chain = theano.shared(test_set_x.value[test_idx:test_idx+20]) # define one step of Gibbs sampling (mf = mean-field) [hid_mf, hid_sample, vis_mf, vis_sample] = rbm.gibbs_vhv(persistent_vis_chain) # the sample at the end of the channel is returned by ``gibbs_1`` as # its second output; note that this is computed as a binomial draw, # therefore it is formed of ints (0 and 1) and therefore needs to # be converted to the same dtype as ``persistent_vis_chain`` vis_sample = T.cast(vis_sample, dtype=theano.config.floatX) # construct the function that implements our persistent chain # we generate the "mean field" activations for plotting and the actual samples for # reinitializing the state of our persistent chain sample_fn = theano.function([], [vis_mf, vis_sample], updates = { persistent_vis_chain:vis_sample}) # sample the RBM, plotting every `plot_every`-th sample; do this # until you plot at least `n_samples` n_samples = 10 plot_every = 1000 for idx in xrange(n_samples): # do `plot_every` intermediate samplings of which we do not care for jdx in xrange(plot_every): vis_mf, vis_sample = sample_fn() # construct image image = PIL.Image.fromarray(tile_raster_images( X = vis_mf, img_shape = (28,28), tile_shape = (10,10), tile_spacing = (1,1) ) ) print ' ... plotting sample ', idx image.save('sample_%i_step_%i.png'%(idx,idx*jdx)) if __name__ == '__main__': test_rbm()