def example_mnist_tap_machine(paysage_path=None,
num_epochs=10,
show_plot=False):
num_hidden_units = 256
batch_size = 100
learning_rate = schedules.power_law_decay(initial=0.1, coefficient=0.1)
(_, _, shuffled_filepath) = \
util.default_paths(paysage_path)
# set up the reader to get minibatches
data = batch.HDFBatch(shuffled_filepath,
'train/images',
batch_size,
transform=batch.binarize_color,
train_fraction=0.95)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = model.Model([vis_layer, hid_layer])
rbm.initialize(data, 'glorot_normal')
perf = fit.ProgressMonitor(
data,
metrics=['ReconstructionError', 'EnergyDistance', 'HeatCapacity'])
opt = optimizers.Gradient(stepsize=learning_rate,
tolerance=1e-4,
ascent=True)
sampler = fit.DrivenSequentialMC.from_batch(rbm, data)
sgd = fit.SGD(rbm,
data,
opt,
num_epochs,
sampler=sampler,
method=fit.tap,
monitor=perf)
# fit the model
print('Training with stochastic gradient ascent using TAP expansion')
sgd.train()
util.show_metrics(rbm, perf)
valid = data.get('validate')
util.show_reconstructions(rbm,
valid,
fit,
show_plot,
n_recon=10,
vertical=False)
util.show_fantasy_particles(rbm, valid, fit, show_plot, n_fantasy=25)
util.show_weights(rbm, show_plot, n_weights=25)
# close the HDF5 store
data.close()
print("Done")
def example_mnist_deep_rbm(paysage_path=None, num_epochs=10, show_plot=False):
num_hidden_units = 500
batch_size = 100
learning_rate = schedules.power_law_decay(initial=0.01, coefficient=0.1)
mc_steps = 1
(_, _, shuffled_filepath) = \
util.default_paths(paysage_path)
# set up the reader to get minibatches
data = batch.HDFBatch(shuffled_filepath,
'train/images',
batch_size,
transform=batch.binarize_color,
train_fraction=0.99)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_1_layer = layers.BernoulliLayer(num_hidden_units)
hid_2_layer = layers.BernoulliLayer(num_hidden_units)
rbm = model.Model([vis_layer, hid_1_layer, hid_2_layer])
rbm.initialize(data)
metrics = [
'ReconstructionError', 'EnergyDistance', 'EnergyGap', 'EnergyZscore',
'HeatCapacity'
]
perf = fit.ProgressMonitor(data, metrics=metrics)
# set up the optimizer and the fit method
opt = optimizers.ADAM(stepsize=learning_rate)
sampler = fit.SequentialMC.from_batch(rbm, data)
cd = fit.SGD(rbm,
data,
opt,
num_epochs,
method=fit.pcd,
sampler=sampler,
mcsteps=mc_steps,
monitor=perf)
# fit the model
print('training with contrastive divergence')
cd.train_layerwise()
# evaluate the model
util.show_metrics(rbm, perf)
valid = data.get('validate')
util.show_reconstructions(rbm, valid, fit, show_plot)
util.show_fantasy_particles(rbm, valid, fit, show_plot)
util.show_weights(rbm, show_plot)
# close the HDF5 store
data.close()
print("Done")
def example_mnist_grbm(paysage_path=None, num_epochs=10, show_plot=False):
num_hidden_units = 500
batch_size = 50
learning_rate = 0.001 # gaussian rbm usually requires smaller learnign rate
mc_steps = 1
(_, _, shuffled_filepath) = \
util.default_paths(paysage_path)
# set up the reader to get minibatches
data = batch.Batch(shuffled_filepath,
'train/images',
batch_size,
transform=transform,
train_fraction=0.99)
# set up the model and initialize the parameters
vis_layer = layers.GaussianLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = model.Model([vis_layer, hid_layer])
rbm.initialize(data)
metrics = [
'ReconstructionError', 'EnergyDistance', 'EnergyGap', 'EnergyZscore'
]
perf = fit.ProgressMonitor(data, metrics=metrics)
opt = optimizers.ADAM(stepsize=learning_rate,
scheduler=optimizers.PowerLawDecay(0.1))
sampler = fit.DrivenSequentialMC.from_batch(rbm, data, method='stochastic')
cd = fit.SGD(rbm,
data,
opt,
num_epochs,
method=fit.pcd,
sampler=sampler,
mcsteps=mc_steps,
monitor=perf)
# fit the model
print('training with contrastive divergence')
cd.train()
# evaluate the model
util.show_metrics(rbm, perf)
util.show_reconstructions(rbm, data.get('validate'), fit, show_plot)
util.show_fantasy_particles(rbm, data.get('validate'), fit, show_plot)
util.show_weights(rbm, show_plot)
# close the HDF5 store
data.close()
print("Done")
def test_gaussian_1D_1mode_train():
# create some example data
num = 10000
mu = 3
sigma = 1
samples = be.randn((num, 1)) * sigma + mu
# set up the reader to get minibatches
batch_size = 100
samples_train, samples_validate = batch.split_tensor(samples, 0.9)
data = batch.Batch({
'train':
batch.InMemoryTable(samples_train, batch_size),
'validate':
batch.InMemoryTable(samples_validate, batch_size)
})
# parameters
learning_rate = schedules.PowerLawDecay(initial=0.1, coefficient=0.1)
mc_steps = 1
num_epochs = 10
num_sample_steps = 100
# set up the model and initialize the parameters
vis_layer = layers.GaussianLayer(1)
hid_layer = layers.OneHotLayer(1)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.initialize(data, method='hinton')
# modify the parameters to shift the initialized model from the data
# this forces it to train
rbm.layers[0].params = layers.ParamsGaussian(
rbm.layers[0].params.loc - 3, rbm.layers[0].params.log_var - 1)
# set up the optimizer and the fit method
opt = optimizers.ADAM(stepsize=learning_rate)
cd = fit.SGD(rbm, data)
# fit the model
print('training with persistent contrastive divergence')
cd.train(opt, num_epochs, method=fit.pcd, mcsteps=mc_steps)
# sample data from the trained model
model_state = \
samplers.SequentialMC.generate_fantasy_state(rbm, num, num_sample_steps)
pts_trained = model_state[0]
percent_error = 10
mu_trained = be.mean(pts_trained)
assert numpy.abs(mu_trained / mu - 1) < (percent_error / 100)
sigma_trained = numpy.sqrt(be.var(pts_trained))
assert numpy.abs(sigma_trained / sigma - 1) < (percent_error / 100)
def run(paysage_path=None, num_epochs=10, show_plot=False):
num_hidden_units = 500
batch_size = 100
learning_rate = schedules.PowerLawDecay(initial=0.001, coefficient=0.1)
mc_steps = 1
(_, _, shuffled_filepath) = \
util.default_paths(paysage_path)
# set up the reader to get minibatches
data = batch.HDFBatch(shuffled_filepath,
'train/images',
batch_size,
transform=pre.binarize_color,
train_fraction=0.99)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.GaussianLayer(num_hidden_units)
hid_layer.set_fixed_params(["loc", "log_var"])
rbm = model.Model([vis_layer, hid_layer])
rbm.initialize(data, method="glorot_normal")
metrics = ['ReconstructionError', 'EnergyDistance', 'EnergyGap',
'EnergyZscore', 'HeatCapacity', 'WeightSparsity', 'WeightSquare']
perf = fit.ProgressMonitor(data, metrics=metrics)
# set up the optimizer and the fit method
opt = optimizers.ADAM(stepsize=learning_rate)
sampler = fit.DrivenSequentialMC.from_batch(rbm, data)
cd = fit.SGD(rbm, data, opt, num_epochs, sampler, method=fit.pcd,
mcsteps=mc_steps, monitor=perf)
# fit the model
print('training with contrastive divergence')
cd.train()
# evaluate the model
util.show_metrics(rbm, perf)
valid = data.get('validate')
util.show_reconstructions(rbm, valid, fit, show_plot,
n_recon=10, vertical=False, num_to_avg=10)
util.show_fantasy_particles(rbm, valid, fit, show_plot, n_fantasy=25)
util.show_weights(rbm, show_plot, n_weights=25)
# close the HDF5 store
data.close()
print("Done")
def run(num_epochs=10, show_plot=False):
num_hidden_units = 256
batch_size = 100
learning_rate = schedules.PowerLawDecay(initial=0.001, coefficient=0.1)
mc_steps = 1
# set up the reader to get minibatches
data = util.create_batch(batch_size,
train_fraction=0.95,
transform=transform)
# set up the model and initialize the parameters
vis_layer = layers.GaussianLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.initialize(data, 'stddev')
rbm.layers[0].params.log_var[:] = \
be.log(0.05*be.ones_like(rbm.layers[0].params.log_var))
opt = optimizers.ADAM(stepsize=learning_rate)
# This example parameter set for TAP uses gradient descent to optimize the
# Gibbs free energy:
tap = fit.TAP(True, 1.0, 0.01, 100, False, 0.9, 0.001, 0.5)
# This example parameter set for TAP uses self-consistent iteration to
# optimize the Gibbs free energy:
#tap = fit.TAP(False, tolerance=0.001, max_iters=100)
sgd = fit.SGD(rbm, data)
sgd.monitor.generator_metrics.append(TAPFreeEnergy())
sgd.monitor.generator_metrics.append(TAPLogLikelihood())
# fit the model
print('Training with stochastic gradient ascent using TAP expansion')
sgd.train(opt, num_epochs, method=tap.tap_update, mcsteps=mc_steps)
util.show_metrics(rbm, sgd.monitor)
valid = data.get('validate')
util.show_reconstructions(rbm,
valid,
show_plot,
n_recon=10,
vertical=False,
num_to_avg=10)
util.show_fantasy_particles(rbm, valid, show_plot, n_fantasy=5)
util.show_weights(rbm, show_plot, n_weights=25)
# close the HDF5 store
data.close()
print("Done")
def run(num_epochs=10, show_plot=False):
num_hidden_units = 100
batch_size = 100
mc_steps = 10
beta_std = 0.6
# set up the reader to get minibatches
with util.create_batch(batch_size,
train_fraction=0.95,
transform=transform) as data:
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units, center=False)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.connections[0].weights.add_penalty(
{'matrix': pen.l2_penalty(0.001)})
rbm.initialize(data, method='pca')
print('training with persistent contrastive divergence')
cd = fit.SGD(rbm, data)
learning_rate = schedules.PowerLawDecay(initial=0.01, coefficient=0.1)
opt = optimizers.ADAM(stepsize=learning_rate)
cd.train(opt, num_epochs, mcsteps=mc_steps, method=fit.pcd)
util.show_metrics(rbm, cd.monitor)
# evaluate the model
valid = data.get('validate')
util.show_reconstructions(rbm,
valid,
show_plot,
n_recon=10,
vertical=False,
num_to_avg=10)
util.show_fantasy_particles(rbm,
valid,
show_plot,
n_fantasy=5,
beta_std=beta_std,
fantasy_steps=100)
util.show_weights(rbm, show_plot, n_weights=100)
print("Done")
return rbm
def example_mnist_tap_machine(paysage_path=None, num_epochs = 10, show_plot=True):
num_hidden_units = 256
batch_size = 100
learning_rate = 0.1
(_, _, shuffled_filepath) = \
util.default_paths(paysage_path)
# set up the reader to get minibatches
data = batch.Batch(shuffled_filepath,
'train/images',
batch_size,
transform=batch.binarize_color,
train_fraction=0.95)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = tap_machine.TAP_rbm([vis_layer, hid_layer], num_persistent_samples=0,
tolerance_EMF=1e-4, max_iters_EMF=25, terms=2)
rbm.initialize(data, 'glorot_normal')
perf = fit.ProgressMonitor(data,
metrics=['ReconstructionError',
'EnergyDistance'])
opt = optimizers.Gradient(stepsize=learning_rate,
scheduler=optimizers.PowerLawDecay(0.1),
tolerance=1e-4,
ascent=True)
sgd = fit.SGD(rbm, data, opt, num_epochs, method=fit.tap, monitor=perf)
# fit the model
print('training with stochastic gradient ascent ')
sgd.train()
util.show_metrics(rbm, perf)
util.show_reconstructions(rbm, data.get('validate'), fit, show_plot)
util.show_fantasy_particles(rbm, data.get('validate'), fit, show_plot)
util.show_weights(rbm, show_plot)
# close the HDF5 store
data.close()
print("Done")
def run(num_epochs=5, show_plot=False):
num_hidden_units = 256
batch_size = 100
learning_rate = schedules.PowerLawDecay(initial=0.1, coefficient=3.0)
mc_steps = 1
# set up the reader to get minibatches
data = util.create_batch(batch_size,
train_fraction=0.95,
transform=transform)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.connections[0].weights.add_penalty(
{'matrix': pen.l1_adaptive_decay_penalty_2(0.00001)})
rbm.initialize(data, 'glorot_normal')
opt = optimizers.Gradient(stepsize=learning_rate, tolerance=1e-4)
tap = fit.TAP(True, 0.1, 0.01, 25, True, 0.5, 0.001, 0.0)
sgd = fit.SGD(rbm, data)
sgd.monitor.generator_metrics.append(TAPLogLikelihood())
sgd.monitor.generator_metrics.append(TAPFreeEnergy())
# fit the model
print('Training with stochastic gradient ascent using TAP expansion')
sgd.train(opt, num_epochs, method=tap.tap_update, mcsteps=mc_steps)
util.show_metrics(rbm, sgd.monitor)
valid = data.get('validate')
util.show_reconstructions(rbm,
valid,
show_plot,
n_recon=10,
vertical=False,
num_to_avg=10)
util.show_fantasy_particles(rbm, valid, show_plot, n_fantasy=5)
util.show_weights(rbm, show_plot, n_weights=25)
# close the HDF5 store
data.close()
print("Done")
def run(num_epochs=1, show_plot=False):
num_hidden_units = 1
batch_size = 100
mc_steps = 10
beta_std = 0.6
# set up the reader to get minibatches
with batch.in_memory_batch(samples, batch_size,
train_fraction=0.95) as data:
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units, center=False)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.connections[0].weights.add_penalty(
{'matrix': pen.l2_penalty(0.001)}) # Add regularization term
rbm.initialize(data, method='hinton') # Initialize weights
cd = fit.SGD(rbm, data)
learning_rate = schedules.PowerLawDecay(initial=0.01,
coefficient=0.1)
opt = optimizers.ADAM(stepsize=learning_rate)
print("Train the model...")
cd.train(opt,
num_epochs,
mcsteps=mc_steps,
method=fit.pcd,
verbose=False)
'''
# write on file KL divergences
reverse_KL_div = [ cd.monitor.memory[i]['ReverseKLDivergence'] for i in range(0,len(cd.monitor.memory)) ]
KL_div = [ cd.monitor.memory[i]['KLDivergence'] for i in range(0,len(cd.monitor.memory)) ]
for i in range(0,len(cd.monitor.memory)):
out_file1.write(str(KL_div[i])+" "+str(reverse_KL_div[i])+"\n")
out_file1.close()
# save weights on file
filename = "results/weights/weights-"+temperature[:-4]+".jpg"
Gprotein_util.show_weights(rbm, show_plot=False, n_weights=8, Filename=filename, random=False)
'''
return rbm
def run(num_epochs=10, show_plot=False):
num_hidden_units = 256
batch_size = 100
learning_rate = schedules.PowerLawDecay(initial=0.001, coefficient=0.1)
mc_steps = 1
# set up the reader to get minibatches
data = util.create_batch(batch_size,
train_fraction=0.95,
transform=transform)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.GaussianLayer(num_hidden_units)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.initialize(data)
# set up the optimizer and the fit method
opt = optimizers.ADAM(stepsize=learning_rate)
cd = fit.SGD(rbm, data)
# fit the model
print('training with contrastive divergence')
cd.train(opt, num_epochs, method=fit.pcd, mcsteps=mc_steps)
# evaluate the model
util.show_metrics(rbm, cd.monitor)
valid = data.get('validate')
util.show_reconstructions(rbm,
valid,
show_plot,
n_recon=10,
vertical=False,
num_to_avg=10)
util.show_fantasy_particles(rbm, valid, show_plot, n_fantasy=5)
util.show_weights(rbm, show_plot, n_weights=25)
# close the HDF5 store
data.close()
print("Done")
def run(num_epochs=20, show_plot=False):
num_hidden_units = 200
batch_size = 100
mc_steps = 10
beta_std = 0.95
# set up the reader to get minibatches
data = util.create_batch(batch_size, train_fraction=0.95, transform=transform)
# set up the model and initialize the parameters
vis_layer = layers.GaussianLayer(data.ncols, center=False)
hid_layer = layers.BernoulliLayer(num_hidden_units, center=True)
hid_layer.set_fixed_params(hid_layer.get_param_names())
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.initialize(data, 'pca', epochs = 500, verbose=True)
print('training with persistent contrastive divergence')
cd = fit.SGD(rbm, data, fantasy_steps=10)
cd.monitor.generator_metrics.append(M.JensenShannonDivergence())
learning_rate = schedules.PowerLawDecay(initial=1e-3, coefficient=5)
opt = optimizers.ADAM(stepsize=learning_rate)
cd.train(opt, num_epochs, method=fit.pcd, mcsteps=mc_steps,
beta_std=beta_std, burn_in=1)
# evaluate the model
util.show_metrics(rbm, cd.monitor)
valid = data.get('validate')
util.show_reconstructions(rbm, valid, show_plot, n_recon=10, vertical=False)
util.show_fantasy_particles(rbm, valid, show_plot, n_fantasy=5)
util.show_weights(rbm, show_plot, n_weights=100)
# close the HDF5 store
data.close()
print("Done")
return rbm
def test_rbm(paysage_path=None):
num_hidden_units = 50
batch_size = 50
num_epochs = 1
learning_rate = schedules.PowerLawDecay(initial=0.01, coefficient=0.1)
mc_steps = 1
if not paysage_path:
paysage_path = os.path.dirname(
os.path.dirname(os.path.abspath(__file__)))
filepath = os.path.join(paysage_path, 'examples', 'mnist', 'mnist.h5')
if not os.path.exists(filepath):
raise IOError(
"{} does not exist. run mnist/download_mnist.py to fetch from the web"
.format(filepath))
shuffled_filepath = os.path.join(paysage_path, 'examples', 'mnist',
'shuffled_mnist.h5')
# shuffle the data
if not os.path.exists(shuffled_filepath):
shuffler = batch.DataShuffler(filepath, shuffled_filepath, complevel=0)
shuffler.shuffle()
# set a seed for the random number generator
be.set_seed()
import pandas
samples = pre.binarize_color(
be.float_tensor(
pandas.read_hdf(shuffled_filepath,
key='train/images').values[:10000]))
samples_train, samples_validate = batch.split_tensor(samples, 0.95)
data = batch.Batch({
'train':
batch.InMemoryTable(samples_train, batch_size),
'validate':
batch.InMemoryTable(samples_validate, batch_size)
})
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.initialize(data)
# obtain initial estimate of the reconstruction error
perf = ProgressMonitor()
untrained_performance = perf.epoch_update(data,
rbm,
store=True,
show=False)
# set up the optimizer and the fit method
opt = optimizers.RMSProp(stepsize=learning_rate)
cd = fit.SGD(rbm, data)
# fit the model
print('training with contrastive divergence')
cd.train(opt, num_epochs, method=fit.pcd, mcsteps=mc_steps)
# obtain an estimate of the reconstruction error after 1 epoch
trained_performance = cd.monitor.memory[-1]
assert (trained_performance['ReconstructionError'] <
untrained_performance['ReconstructionError']), \
"Reconstruction error did not decrease"
# close the HDF5 store
data.close()
def test_rbm(paysage_path=None):
num_hidden_units = 50
batch_size = 50
num_epochs = 1
learning_rate = schedules.PowerLawDecay(initial=0.01, coefficient=0.1)
mc_steps = 1
if not paysage_path:
paysage_path = os.path.dirname(
os.path.dirname(os.path.abspath(__file__)))
filepath = os.path.join(paysage_path, 'mnist', 'mnist.h5')
if not os.path.exists(filepath):
raise IOError(
"{} does not exist. run mnist/download_mnist.py to fetch from the web"
.format(filepath))
shuffled_filepath = os.path.join(paysage_path, 'mnist',
'shuffled_mnist.h5')
# shuffle the data
if not os.path.exists(shuffled_filepath):
shuffler = batch.DataShuffler(filepath, shuffled_filepath, complevel=0)
shuffler.shuffle()
# set a seed for the random number generator
be.set_seed()
# set up the reader to get minibatches
data = batch.HDFBatch(shuffled_filepath,
'train/images',
batch_size,
transform=pre.binarize_color,
train_fraction=0.99)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = model.Model([vis_layer, hid_layer])
rbm.initialize(data)
# obtain initial estimate of the reconstruction error
perf = fit.ProgressMonitor(data, metrics=['ReconstructionError'])
untrained_performance = perf.check_progress(rbm)
# set up the optimizer and the fit method
opt = optimizers.RMSProp(stepsize=learning_rate)
sampler = fit.DrivenSequentialMC.from_batch(rbm, data)
cd = fit.SGD(rbm,
data,
opt,
num_epochs,
sampler,
method=fit.pcd,
mcsteps=mc_steps,
monitor=perf)
# fit the model
print('training with contrastive divergence')
cd.train()
# obtain an estimate of the reconstruction error after 1 epoch
trained_performance = perf.check_progress(rbm)
assert (trained_performance['ReconstructionError'] <
untrained_performance['ReconstructionError']), \
"Reconstruction error did not decrease"
# close the HDF5 store
data.close()
def test_tap_machine(paysage_path=None):
num_hidden_units = 10
batch_size = 50
num_epochs = 1
learning_rate = 0.01
if not paysage_path:
paysage_path = os.path.dirname(
os.path.dirname(os.path.abspath(__file__)))
filepath = os.path.join(paysage_path, 'mnist', 'mnist.h5')
if not os.path.exists(filepath):
raise IOError(
"{} does not exist. run mnist/download_mnist.py to fetch from the web"
.format(filepath))
shuffled_filepath = os.path.join(paysage_path, 'mnist',
'shuffled_mnist.h5')
# shuffle the data
if not os.path.exists(shuffled_filepath):
shuffler = batch.DataShuffler(filepath, shuffled_filepath, complevel=0)
shuffler.shuffle()
# set a seed for the random number generator
be.set_seed()
# set up the reader to get minibatches
data = batch.Batch(shuffled_filepath,
'train/images',
batch_size,
transform=batch.binarize_color,
train_fraction=0.1)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = tap_machine.TAP_rbm([vis_layer, hid_layer],
tolerance_EMF=1e-2,
max_iters_EMF=50)
rbm.initialize(data)
# obtain initial estimate of the reconstruction error
perf = fit.ProgressMonitor(data, metrics=['ReconstructionError'])
untrained_performance = perf.check_progress(rbm)
# set up the optimizer and the fit method
opt = optimizers.Gradient(stepsize=learning_rate,
scheduler=optimizers.PowerLawDecay(0.1),
tolerance=1e-3,
ascent=True)
solver = fit.SGD(rbm, data, opt, num_epochs, method=fit.tap, monitor=perf)
# fit the model
print('training with stochastic gradient ascent')
solver.train()
# obtain an estimate of the reconstruction error after 1 epoch
trained_performance = perf.check_progress(rbm)
assert (trained_performance['ReconstructionError'] <
untrained_performance['ReconstructionError']), \
"Reconstruction error did not decrease"
# close the HDF5 store
data.close()
def run(paysage_path=None, num_epochs=10, show_plot=False):
num_hidden_units = 256
batch_size = 100
learning_rate = schedules.PowerLawDecay(initial=0.01, coefficient=0.1)
mc_steps = 1
(_, _, shuffled_filepath) = \
util.default_paths(paysage_path)
# set up the reader to get minibatches
import pandas
data = batch.InMemoryBatch(pre.binarize_color(
be.float_tensor(
pandas.read_hdf(shuffled_filepath,
key='train/images').as_matrix())),
batch_size,
train_fraction=0.95)
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = model.Model([vis_layer, hid_layer])
rbm.weights[0].add_penalty({'matrix': pen.l2_penalty(0.001)})
rbm.initialize(data, method='glorot_normal')
metrics = [
'ReconstructionError', 'EnergyDistance', 'EnergyGap', 'EnergyZscore',
'HeatCapacity', 'WeightSparsity', 'WeightSquare'
]
perf = fit.ProgressMonitor(data, metrics=metrics)
# set up the optimizer and the fit method
opt = optimizers.ADAM(stepsize=learning_rate)
sampler = fit.DrivenSequentialMC.from_batch(rbm, data)
cd = fit.SGD(rbm,
data,
opt,
num_epochs,
sampler,
method=fit.pcd,
mcsteps=mc_steps,
monitor=perf)
# fit the model
print('training with contrastive divergence')
cd.train()
# evaluate the model
util.show_metrics(rbm, perf)
valid = data.get('validate')
util.show_reconstructions(rbm,
valid,
fit,
show_plot,
n_recon=10,
vertical=False,
num_to_avg=10)
util.show_fantasy_particles(rbm, valid, fit, show_plot, n_fantasy=25)
util.show_weights(rbm, show_plot, n_weights=25)
# close the HDF5 store
data.close()
print("Done")
def test_tap_machine(paysage_path=None):
num_hidden_units = 10
batch_size = 100
num_epochs = 5
learning_rate = schedules.PowerLawDecay(initial=0.1, coefficient=1.0)
if not paysage_path:
paysage_path = os.path.dirname(
os.path.dirname(os.path.abspath(__file__)))
filepath = os.path.join(paysage_path, 'examples', 'mnist', 'mnist.h5')
if not os.path.exists(filepath):
raise IOError(
"{} does not exist. run mnist/download_mnist.py to fetch from the web"
.format(filepath))
shuffled_filepath = os.path.join(paysage_path, 'examples', 'mnist',
'shuffled_mnist.h5')
# shuffle the data
if not os.path.exists(shuffled_filepath):
shuffler = batch.DataShuffler(filepath, shuffled_filepath, complevel=0)
shuffler.shuffle()
# set a seed for the random number generator
be.set_seed()
# set up the reader to get minibatches
samples = pre.binarize_color(
be.float_tensor(
pandas.read_hdf(shuffled_filepath,
key='train/images').as_matrix()[:10000]))
samples_train, samples_validate = batch.split_tensor(samples, 0.95)
data = batch.Batch({
'train':
batch.InMemoryTable(samples_train, batch_size),
'validate':
batch.InMemoryTable(samples_validate, batch_size)
})
# set up the model and initialize the parameters
vis_layer = layers.BernoulliLayer(data.ncols)
hid_layer = layers.BernoulliLayer(num_hidden_units)
rbm = BoltzmannMachine([vis_layer, hid_layer])
rbm.initialize(data)
# obtain initial estimate of the reconstruction error
perf = ProgressMonitor(generator_metrics = \
[ReconstructionError(), TAPLogLikelihood(10), TAPFreeEnergy(10)])
untrained_performance = perf.epoch_update(data,
rbm,
store=True,
show=False)
# set up the optimizer and the fit method
opt = optimizers.Gradient(stepsize=learning_rate, tolerance=1e-5)
tap = fit.TAP(True, 0.1, 0.01, 25, True, 0.5, 0.001, 0.0)
solver = fit.SGD(rbm, data)
solver.monitor.generator_metrics.append(TAPLogLikelihood(10))
solver.monitor.generator_metrics.append(TAPFreeEnergy(10))
# fit the model
print('training with stochastic gradient ascent')
solver.train(opt, num_epochs, method=tap.tap_update)
# obtain an estimate of the reconstruction error after 1 epoch
trained_performance = solver.monitor.memory[-1]
assert (trained_performance['TAPLogLikelihood'] >
untrained_performance['TAPLogLikelihood']), \
"TAP log-likelihood did not increase"
assert (trained_performance['ReconstructionError'] <
untrained_performance['ReconstructionError']), \
"Reconstruction error did not decrease"
# close the HDF5 store
data.close()
