def associate_data2dataADBN_Finetune(cache=False, train_further=False):
print "Testing Associative RBM which tries to learn even-oddness of numbers"
f = open('adbn_errors.txt', 'w')
# project set-up
proj_name = 'ADBN_digits'
data_manager = store.StorageManager(proj_name, log=True)
shape = 28
train_n = 10000
test_n = 1000
# Load mnist hand digits, class label is already set to binary
dataset = m_loader.load_digits(n=[train_n, 0, test_n],
digits=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
pre={'binary_label': True})
tr_x, tr_y = dataset[0]
te_x, te_y = dataset[2]
tr_x01 = m_loader.sample_image(tr_y)
te_x01 = m_loader.sample_image(te_y)
clf = SimpleClassifier('logistic', te_x01.get_value(), te_y.eval())
configs = get_adbns()
for a in xrange(10):
for n, config in enumerate(configs):
t = 'l{}_r{}_t{}'.format(config.left_dbn.topology, config.right_dbn.topology, config.n_association)
f.write('{}:{}:'.format(a, t))
brain_c = associative_dbn.AssociativeDBN(config=config,
data_manager=store.StorageManager('{}/{}'.format(proj_name, n),
log=False))
brain_c.train(tr_x, tr_x01, cache=True, train_further=True)
recon = brain_c.recall(tr_x,
associate_steps=10,
recall_steps=0,
img_name='{}_{}'.format(a, t))
error = clf.get_score(recon, tr_y.eval())
print error
f.write('{}, '.format(error))
for i in xrange(0, 10):
brain_c.fine_tune(tr_x, tr_x01, epochs=1)
for y_type in ['active_h', 'v_noisy_active_h', 'zero']:
recon = brain_c.recall(tr_x,
associate_steps=10,
recall_steps=0,
img_name='{}_{}_{}_{}'.format(a, t, y_type, i),
y_type=y_type)
error = clf.get_score(recon, tr_y.eval())
print error
f.write('{}, '.format(error))
f.write('\n')
f.close()
def associate_data2dataADBN(cache=False, train_further=True):
print "Testing Associative RBM which tries to learn even-oddness of numbers"
f = open('adbnlog.txt', 'a')
# project set-up
data_manager = store.StorageManager('AssociativeDBN_digits', log=False)
shape = 28
train_n = 10000
test_n = 1000
# Load mnist hand digits, class label is already set to binary
dataset = m_loader.load_digits(n=[train_n, 100, test_n],
digits=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
pre={'binary_label': True})
tr_x, tr_y = dataset[0]
te_x, te_y = dataset[2]
tr_x01 = m_loader.sample_image(tr_y)
clf = SimpleClassifier('logistic', te_x.get_value(), te_y.eval())
for dropout in [True, False]:
for sc in [True, False]:
for lr in [0.001, 0.0001, 0.00005, 0.00001]:
for n in [100, 250, 500]:
config = get_brain_model_AssociativeDBNConfig(28, data_manager=data_manager)
config.top_rbm.train_params.learning_rate = lr
config.top_rbm.train_params.sparsity_constraint = sc
config.top_rbm.train_params.dropout = dropout
config.n_association = n
adbn = associative_dbn.AssociativeDBN(config=config, data_manager=data_manager)
brain_c = adbn
f.write(str(brain_c.association_layer) + "\n")
errors = []
for i in xrange(0, 5):
f.write("Epoch %d \n" % (i * 10))
brain_c.train(tr_x, tr_x01,
cache=[[True, True, True], [True, True, True], False],
train_further=[[False, True, True], [False, True, True], True])
if i == 0:
# Reconstruction
recon_right = brain_c.dbn_left.reconstruct(tr_x, k=10, plot_every=1, plot_n=100,
img_name='adbn_left_recon_{}'.format(shape))
recon_left = brain_c.dbn_right.reconstruct(tr_x01, k=10, plot_every=1, plot_n=100,
img_name='adbn_right_recon_{}'.format(shape))
for j in [5, 10]:
recon_x = brain_c.recall(te_x, associate_steps=j, recall_steps=0,
img_name='adbn_child_recon_{}1'.format(shape), y_type='active_h')
error = clf.get_score(recon_x, te_y.eval())
f.write("active_h %f\n" % error)
recon_x = brain_c.recall(te_x, associate_steps=j, recall_steps=0,
img_name='adbn_child_recon_{}2'.format(shape),
y_type='v_noisy_active_h')
error = clf.get_score(recon_x, te_y.eval())
f.write("v_noisy_active_h %f\n" % error)
f.close()
def train_kanade():
print "Testing RBM"
data_manager = store.StorageManager('Kanade/SimpleRBMTest')
# Load mnist hand digits
datasets = loader.load_kanade(n=500,
set_name='25_25',
emotions=['happy', 'sadness'],
pre={'scale2unit': True})
train_x, train_y = datasets[0]
sparsity_constraint = True
# Initialise the RBM and training parameters
tr = rbm_config.TrainParam(learning_rate=0.0001,
momentum_type=rbm_config.NESTEROV,
momentum=0.9,
weight_decay=0.001,
sparsity_constraint=sparsity_constraint,
sparsity_target=0.01,
sparsity_cost=1,
sparsity_decay=0.9,
dropout=True,
epochs=100)
n_visible = train_x.get_value().shape[1]
n_hidden = 500
config = rbm_config.RBMConfig(
v_n=n_visible,
v2_n=n_visible,
h_n=n_hidden,
v_unit=rbm_units.GaussianVisibleUnit,
associative=False,
cd_type=rbm_config.CLASSICAL,
cd_steps=1,
train_params=tr,
progress_logger=rbm_logger.ProgressLogger(img_shape=(25, 25)))
rbm = RBM(config)
print "... initialised RBM"
# Train RBM
rbm.train(train_x)
# Test RBM
rbm.reconstruct(train_x, k=5, plot_n=10, plot_every=1)
# Store Parameters
data_manager.persist(rbm)
def test_rbm():
print "Testing RBM"
data_manager = store.StorageManager('TestRBM')
# Load Cohn Kanade dataset
datasets = loader.load_kanade(pre={'scale': True}, n=100, set_name='sharp_equi25_25')
train_set_x, train_set_y = datasets[0]
test_set_x, test_set_y = datasets[2]
# Initilise the RBM and training parameters
tr = TrainParam(learning_rate=0.0001,
momentum_type=NESTEROV,
momentum=0.5,
weight_decay=0.0001,
sparsity_constraint=True,
sparsity_target=0.01,
sparsity_cost=0.1,
sparsity_decay=0.9,
dropout=True,
dropout_rate=0.5,
batch_size=10,
epochs=10)
n_visible = train_set_x.get_value(borrow=True).shape[1]
n_hidden = 10
config = RBMConfig()
config.v_n = n_visible
config.h_n = n_hidden
config.v_unit = rbm_units.GaussianVisibleUnit
# config.h_unit = rbm_units.ReLUnit
config.progress_logger = ProgressLogger(img_shape=(25, 25))
config.train_params = tr
rbm = RBM(config)
print "... initialised RBM"
load = store.retrieve_object(str(rbm))
if load:
rbm = load
for i in xrange(0, 1):
# Train RBM
rbm.train(train_set_x)
data_manager.persist(rbm)
# Test RBM Reconstruction via Linear Classifier
clf = SimpleClassifier(classifier='logistic', train_x=train_set_x, train_y=train_set_y)
recon_te = rbm.reconstruct(test_set_x, k=1, plot_n=100, plot_every=1,img_name='recon_te_{}.png'.format(i))
print 'Original Score: {}'.format(clf.get_score(test_set_x, test_set_y))
print 'Recon Score: {}'.format(clf.get_score(recon_te, test_set_y.eval()))
def test_generative_dbn():
manager = store.StorageManager('fine_tune')
shape = 28
train_x = get_data(shape)
# Initialise RBM parameters
dbn = get_dbn_model(manager, shape)
print "... initialised dbn"
print '... pre-training the model'
start_time = time.clock()
dbn.pretrain(train_x, cache=[True, True], train_further=[True, True])
# dbn.pretrain(train_x, cache=False)
end_time = time.clock()
print >> sys.stderr, ('The pretraining code for file ' +
os.path.split(__file__)[1] + ' ran for %.2fm' %
((end_time - start_time) / 60.))
# Sample from top layer to generate data
sample_n = 1000
sampled = dbn.sample(sample_n, k=100, rand_type='noisy_mean')
k_loader.save_faces(sampled,
tile=(sample_n / 10, 10),
img_name="sampled.png",
img_shape=(shape, shape))
dbn.reconstruct(train_x,
k=1,
plot_every=1,
plot_n=100,
img_name='dbn_recon')
for i in xrange(0, 1):
dbn.fine_tune(train_x, epochs=1)
sampled = dbn.sample(sample_n, k=100, rand_type='noisy_mean')
k_loader.save_faces(sampled,
tile=(sample_n / 10, 10),
img_name=("sampled_fine_tuned%d.png" % i),
img_shape=(shape, shape))
dbn.reconstruct(train_x,
k=1,
plot_every=1,
plot_n=100,
img_name=('dbn_recon_fine_tune%d' % i))
def associate_data2data(cache=False, train_further=True):
print "Testing Associative RBM which tries to learn even-oddness of numbers"
# project set-up
data_manager = store.StorageManager('EvenOddP', log=True)
train_n = 10000
test_n = 1000
# Load mnist hand digits, class label is already set to binary
dataset = m_loader.load_digits(n=[train_n, 100, test_n],
digits=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
pre={'binary_label': True})
tr_x, tr_y = dataset[0]
te_x, te_y = dataset[2]
tr_x01 = m_loader.sample_image(tr_y)
te_x01 = m_loader.sample_image(te_y)
ones = m_loader.load_digits(n=[test_n, 0, 0], digits=[1])[0][0]
zeroes = m_loader.load_digits(n=[test_n, 0, 0], digits=[0])[0][0]
concat1 = theano.function([], T.concatenate([tr_x, tr_x01], axis=1))()
# concat2 = theano.function([], T.concatenate([tr_x01, tr_x], axis=1))()
# c = np.concatenate([concat1, concat2], axis=0)
# np.random.shuffle(c)
# tr_concat_x = theano.shared(c, name='tr_concat_x')
tr_concat_x = theano.shared(concat1, name='tr_concat_x')
tr = TrainParam(learning_rate=0.001,
momentum_type=NESTEROV,
momentum=0.5,
weight_decay=0.1,
sparsity_constraint=True,
sparsity_target=0.1,
sparsity_decay=0.9,
sparsity_cost=0.1,
dropout=True,
dropout_rate=0.5,
epochs=1)
# Even odd test
k = 1
n_visible = 784 * 2
n_visible2 = 0
n_hidden = 300
print "number of hidden nodes: %d" % n_hidden
config = RBMConfig(v_n=n_visible,
v2_n=n_visible2,
h_n=n_hidden,
cd_type=CLASSICAL,
cd_steps=k,
train_params=tr,
progress_logger=ProgressLogger(img_shape=(28 * 2, 28)))
rbm = RBM(config=config)
# Load RBM (test)
loaded = store.retrieve_object(str(rbm))
if loaded and cache:
rbm = loaded
print "... loaded precomputed rbm"
errors = []
for i in xrange(0, 10):
# Train RBM
if not loaded or train_further:
rbm.train(tr_concat_x)
# Save RBM
data_manager.persist(rbm)
# Reconstruct using RBM
recon_x = rbm.reconstruct_association_opt(te_x, k=10, bit_p=0)
clf = SimpleClassifier('logistic', te_x.get_value(), te_y.eval())
orig = te_y.eval()
error = clf.get_score(recon_x, orig)
print error
errors.append(error)
print errors
def associate_data2dataJDBN(cache=False, train_further=False):
print "Testing Associative RBM which tries to learn even-oddness of numbers"
f = open('errors.txt', 'w')
# project set-up
proj_name = 'JDBN_digits'
data_manager = store.StorageManager(proj_name, log=False)
shape = 28
train_n = 10000
test_n = 1000
# Load mnist hand digits, class label is already set to binary
dataset = m_loader.load_digits(n=[train_n, 0, test_n],
digits=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
pre={'binary_label': True})
tr_x, tr_y = dataset[0]
te_x, te_y = dataset[2]
tr_x01 = m_loader.sample_image(tr_y)
te_x01 = m_loader.sample_image(te_y)
ones = m_loader.load_digits(n=[test_n, 0, 0], digits=[1])[0][0]
zeroes = m_loader.load_digits(n=[test_n, 0, 0], digits=[0])[0][0]
tr_X = theano.shared(np.concatenate([tr_x.get_value(), tr_x01.get_value()], axis=1))
initial_y = np.random.binomial(n=1, p=0.0, size=te_x.get_value(True).shape).astype(t_float_x)
initial_y_uni = np.random.uniform(low=0, high=1, size=te_x.get_value(True).shape).astype(t_float_x)
initial_y_bi001 = np.random.binomial(n=1, p=0.01, size=te_x.get_value(True).shape).astype(t_float_x)
initial_y_bi01 = np.random.binomial(n=1, p=0.1, size=te_x.get_value(True).shape).astype(t_float_x)
te_X = theano.shared(np.concatenate([te_x.get_value(), initial_y], axis=1))
te_X2 = theano.shared(np.concatenate([te_x.get_value(), initial_y_uni], axis=1))
te_X3 = theano.shared(np.concatenate([te_x.get_value(), initial_y_bi01], axis=1))
te_X4 = theano.shared(np.concatenate([te_x.get_value(), initial_y_bi001], axis=1))
clf = SimpleClassifier('logistic', te_x01.get_value(), te_y.eval())
configs = []
for h_n in [100, 250, 500]:
config1 = get_brain_model_JointDBNConfig(shape, data_manager)
config1.topology = [784 * 2, h_n, h_n]
config1.rbm_configs[0].h_n = h_n
config1.rbm_configs[1].v_n = h_n
config1.rbm_configs[1].h_n = h_n
configs.append(config1)
for h_n in [100, 250, 500]:
config1 = get_brain_model_JointDBNConfig(shape, data_manager)
config1.topology = [784 * 2, h_n, h_n * 2]
config1.rbm_configs[0].h_n = h_n
config1.rbm_configs[1].v_n = h_n
config1.rbm_configs[1].h_n = h_n * 2
configs.append(config1)
for h_n in [100, 250, 500]:
config1 = get_brain_model_JointDBNConfig(shape, data_manager)
config1.topology = [784 * 2, h_n * 2, h_n]
config1.rbm_configs[0].h_n = h_n * 2
config1.rbm_configs[1].v_n = h_n * 2
config1.rbm_configs[1].h_n = h_n
configs.append(config1)
for a in xrange(10):
for config in configs:
f.write('{}:{}:'.format(a, config.topology))
brain_c = DBN.DBN(config=config)
brain_c.pretrain(tr_X, cache=[True, True], train_further=[True, True])
recon_x = brain_c.reconstruct(te_X, k=1, plot_n=100, img_name='{}_{}_recon'.format(a, config.topology))
recon = recon_x[:, 784:]
error = clf.get_score(recon, te_y.eval())
print error
f.write('{}, '.format(error))
for i in xrange(0, 5):
brain_c.fine_tune(tr_X)
recon_x = brain_c.reconstruct(te_X, k=1, plot_n=100,
img_name=('{}_{}_recon_fine_tune{}'.format(a, config.topology, i)))
recon = recon_x[:, 784:]
error = clf.get_score(recon, te_y.eval())
print error
f.write('{}, '.format(error))
recon_x = brain_c.reconstruct(te_X4, k=1, plot_n=100,
img_name=('{}_{}_recon_fine_tune_2_{}'.format(a, config.topology, i)))
recon = recon_x[:, 784:]
error = clf.get_score(recon, te_y.eval())
print error
f.write('{}, '.format(error))
f.write('\n')
f.close()
def associate_data2dataDBN(cache=False):
print "Testing Joint DBN which tries to learn even-oddness of numbers"
# project set-up
data_manager = store.StorageManager('associative_dbn_test', log=True)
# Load mnist hand digits, class label is already set to binary
train, valid, test = m_loader.load_digits(n=[500, 100, 100], digits=[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
pre={'binary_label': True})
train_x, train_y = train
test_x, test_y = test
train_x01 = m_loader.sample_image(train_y)
dataset01 = m_loader.load_digits(n=[500, 100, 100], digits=[0, 1])
# Initialise RBM parameters
# fixed base train param
base_tr = RBM.TrainParam(learning_rate=0.01,
momentum_type=RBM.CLASSICAL,
momentum=0.5,
weight_decay=0.0005,
sparsity_constraint=False,
epochs=20)
# top layer parameters
tr = RBM.TrainParam(learning_rate=0.1,
find_learning_rate=True,
momentum_type=RBM.NESTEROV,
momentum=0.5,
weight_decay=0.001,
sparsity_constraint=False,
epochs=20)
tr_top = RBM.TrainParam(learning_rate=0.1,
find_learning_rate=True,
momentum_type=RBM.CLASSICAL,
momentum=0.5,
weight_decay=0.001,
sparsity_constraint=False,
epochs=20)
# Layer 1
# Layer 2
# Layer 3
topology = [784, 500, 500, 100]
config = associative_dbn.DefaultADBNConfig()
config.topology_left = [784, 500, 500, 100]
config.topology_right = [784, 500, 500, 100]
config.reuse_dbn = False
config.top_rbm_params = tr_top
config.base_rbm_params = [base_tr, tr, tr]
for cd_type in [RBM.CLASSICAL, RBM.PERSISTENT]:
for n_ass in [100, 250, 500, 750, 1000]:
config.n_association = n_ass
config.top_cd_type = cd_type
# Construct DBN
assoc_dbn = associative_dbn.AssociativeDBN(config=config, data_manager=data_manager)
# Train
assoc_dbn.train(train_x, train_x01, cache=cache, optimise=True)
for n_recall in [1, 3, 5, 7, 10]:
for n_think in [0, 1, 3, 5, 7, 10]: # 1, 3, 5, 7, 10]:
# Reconstruct
sampled = assoc_dbn.recall(test_x, n_recall, n_think)
# Sample from top layer to generate data
sample_n = 100
utils.save_images(sampled, image_name='reconstruced_{}_{}_{}.png'.format(n_ass, n_recall, n_think),
shape=(sample_n / 10, 10))
dataset01[2] = (theano.shared(sampled), test_y)
def KanadeAssociativeRBM(cache=False, train_further=False):
print "Testing Associative RBM which tries to learn the ID map "
# print "Testing Associative RBM which tries to learn the following mapping: {anger, saddness, disgust} -> {sadness}, {contempt, happy, surprise} -> {happy}"
# project set-up
data_manager = store.StorageManager('Kanade/OptMFSparse0.01RBMTest',
log=True)
# data_manager = store.StorageManager('Kanade/OptAssociativeRBMTest', log=True)
shape = 25
dataset_name = 'sharp_equi{}_{}'.format(shape, shape)
# Load kanade database
mapping = None # id map
# mapping = {'anger': 'sadness', 'contempt': 'happy', 'disgust': 'sadness', 'fear': 'sadness', 'happy': 'happy',
# 'sadness': 'sadness', 'surprise': 'happy'}
train, valid, test = loader.load_kanade(pre={'scale': True},
set_name=dataset_name)
train_x, train_y = train
test_x, test_y = test
# Sample associated image
train_x_mapped, train_y_mapped = loader.sample_image(train_y,
mapping=mapping,
pre={'scale': True},
set_name=dataset_name)
test_x_mapped, test_y_mapped = loader.sample_image(test_y,
mapping=mapping,
pre={'scale': True},
set_name=dataset_name)
# Concatenate images
concat1 = T.concatenate([train_x, train_x_mapped], axis=1)
# concat2 = T.concatenate([train_x_mapped, train_x], axis=1)
# concat = T.concatenate([concat1, concat2], axis=0)
# train_tX = theano.function([], concat)()
train_tX = theano.function([], concat1)()
train_X = theano.shared(train_tX)
# Train classifier to be used for classifying reconstruction associated image layer
# mapped_data = loader.load_kanade(#emotions=['sadness', 'happy'],
# pre={'scale': True},
# set_name=dataset_name) # Target Image
# clf_orig = SimpleClassifier('logistic', mapped_data[0][0], mapped_data[0][1])
clf_orig = SimpleClassifier('logistic', train_x, train_y)
# Initialise RBM
tr = rbm_config.TrainParam(learning_rate=0.0001,
momentum_type=rbm_config.NESTEROV,
momentum=0.9,
weight_decay=0.0001,
sparsity_constraint=True,
sparsity_target=0.01,
sparsity_cost=100,
sparsity_decay=0.9,
batch_size=10,
epochs=10)
n_visible = shape * shape * 2
n_hidden = 500
config = rbm_config.RBMConfig()
config.v_n = n_visible
config.h_n = n_hidden
config.v_unit = rbm_units.GaussianVisibleUnit
# config.h_unit = rbm_units.ReLUnit
config.progress_logger = rbm_logger.ProgressLogger(img_shape=(shape * 2,
shape))
config.train_params = tr
rbm = RBM(config)
print "... initialised RBM"
# Load RBM (test)
loaded = data_manager.retrieve(str(rbm))
if loaded:
rbm = loaded
else:
rbm.set_initial_hidden_bias()
rbm.set_hidden_mean_activity(train_X)
# Train RBM - learn joint distribution
# rbm.pretrain_lr(train_x, train_x01)
for i in xrange(0, 10):
if not cache or train_further:
rbm.train(train_X)
data_manager.persist(rbm)
print "... reconstruction of associated images"
# Get reconstruction with train data to get 'mapped' images to train classifiers on
reconstruction = rbm.reconstruct(train_X,
1,
plot_n=100,
plot_every=1,
img_name='recon_train')
reconstruct_assoc_part = reconstruction[:, (shape**2):]
# Get associated images of test data
nsamples = np.random.normal(0, 1,
test_x.get_value(True).shape).astype(
np.float32)
initial_y = theano.shared(nsamples, name='initial_y')
utils.save_images(nsamples[0:100], 'initialisation.png', (10, 10),
(25, 25))
test_x_associated = rbm.reconstruct_association_opt(
test_x,
initial_y,
5,
0.,
plot_n=100,
plot_every=1,
img_name='recon_test_gibbs')
mf_recon = rbm.mean_field_inference_opt(test_x,
y=initial_y,
sample=False,
k=10,
img_name='recon_test_mf_raw')
# Concatenate images
test_MFX = theano.function([], T.concatenate([test_x, mf_recon],
axis=1))()
test_MF = theano.shared(test_MFX)
reconstruction = rbm.reconstruct(test_MF,
1,
plot_n=100,
plot_every=1,
img_name='recon_test_mf_recon')
mf_recon = reconstruction[:, (shape**2):]
print "... reconstructed"
# Classify the reconstructions
# 1. Train classifier on original images
score_orig = clf_orig.get_score(test_x_associated,
test_y_mapped.eval())
score_orig_mf = clf_orig.get_score(test_x_associated,
test_y_mapped.eval())
# 2. Train classifier on reconstructed images
clf_recon = SimpleClassifier('logistic', reconstruct_assoc_part,
train_y_mapped.eval())
score_retrain = clf_recon.get_score(test_x_associated,
test_y_mapped.eval())
score_retrain_mf = clf_recon.get_score(mf_recon, test_y_mapped.eval())
out_msg = '{} (orig, retrain):{},{}'.format(rbm, score_orig,
score_retrain)
out_msg2 = '{} (orig, retrain):{},{}'.format(rbm, score_orig_mf,
score_retrain_mf)
print out_msg
print out_msg2
def associate_data2dataDBN(cache=False):
print "Testing Associative DBN which tries to learn even-oddness of numbers"
# project set-up
data_manager = store.StorageManager('Kanade/associative_dbn_test',
log=True)
# Load mnist hand digits, class label is already set to binary
dataset = loader.load_kanade(n=500,
emotions=['anger', 'sadness', 'happy'],
pre={'scale2unit': True})
train_x, train_y = dataset
train_x01 = loader.sample_image(train_y)
dataset01 = loader.load_kanade(n=500)
# Initialise RBM parameters
# fixed base train param
base_tr = RBM.TrainParam(learning_rate=0.001,
momentum_type=RBM.CLASSICAL,
momentum=0.5,
weight_decay=0.0005,
sparsity_constraint=False,
epochs=20)
# top layer parameters
tr = RBM.TrainParam(
learning_rate=0.001,
# find_learning_rate=True,
momentum_type=RBM.NESTEROV,
momentum=0.5,
weight_decay=0.001,
sparsity_constraint=False,
epochs=20)
tr_top = RBM.TrainParam(
learning_rate=0.001,
# find_learning_rate=True,
momentum_type=RBM.CLASSICAL,
momentum=0.5,
weight_decay=0.001,
sparsity_constraint=False,
epochs=20)
# Layer 1
# Layer 2
# Layer 3
# topology = [784, 500, 500, 100]
config = associative_dbn.DefaultADBNConfig()
config.topology_left = [625, 500, 500, 100]
config.topology_right = [625, 500, 500, 100]
config.reuse_dbn = False
config.top_rbm_params = tr_top
config.base_rbm_params = [base_tr, tr, tr]
count = 0
for cd_type in [RBM.CLASSICAL, RBM.PERSISTENT]:
for n_ass in [100, 250, 500, 750, 1000]:
config.n_association = n_ass
config.top_cd_type = cd_type
# Construct DBN
ass_dbn = associative_dbn.AssociativeDBN(config=config,
data_manager=data_manager)
# Train
for trainN in xrange(0, 5):
ass_dbn.train(train_x, train_x01, cache=cache)
for n_recall in [1, 3, 10]:
for n_think in [0, 1, 3, 5, 10]: # 1, 3, 5, 7, 10]:
# Reconstruct
sampled = ass_dbn.recall(train_x, n_recall, n_think)
# Sample from top layer to generate data
sample_n = 100
utils.save_images(
sampled,
image_name='{}_reconstruced_{}_{}_{}.png'.format(
count, n_ass, n_recall, n_think),
shape=(sample_n / 10, 10),
img_shape=(25, 25))
count += 1
def KanadeJointDBN(cache=False):
print "Testing JointDBN which tries to learn id map association"
# project set-up
data_manager = store.StorageManager('Kanade/JointDBN', log=True)
shape = 25
dataset_name = 'sharp_equi{}_{}'.format(shape, shape)
preprocessing = {'scale': True}
# Load kanade database
mapping = None
# mapping = {'anger': 'sadness',
# 'contempt': 'happy',
# 'disgust': 'sadness',
# 'fear': 'sadness',
# 'happy': 'happy',
# 'sadness': 'sadness',
# 'surprise': 'happy'}
dataset = loader.load_kanade( # n=3000,
pre=preprocessing, set_name=dataset_name)
mapped_dataset = loader.load_kanade( # n=3000,
# emotions=['sadness', 'happy'],
pre=preprocessing,
set_name=dataset_name) # Target Image
train, valid, test = dataset
train_x, train_y = train
test_x, test_y = test
# Sample associated image
train_x_ass, train_y_ass = loader.sample_image(train_y,
mapping=mapping,
pre=preprocessing,
set_name=dataset_name)
test_x_ass, test_y_ass = loader.sample_image(test_y,
mapping=mapping,
pre=preprocessing,
set_name=dataset_name)
# Initialise RBM parameters
base_tr = rbm_config.TrainParam(learning_rate=0.0001,
momentum_type=rbm_config.NESTEROV,
momentum=0.9,
weight_decay=0.0001,
sparsity_constraint=False,
sparsity_target=0.00001,
sparsity_decay=0.9,
sparsity_cost=10000,
epochs=100,
batch_size=10)
rest_tr = rbm_config.TrainParam(learning_rate=0.0001,
momentum_type=rbm_config.CLASSICAL,
momentum=0.5,
weight_decay=0.01,
epochs=100,
batch_size=10)
# Layer 1
# Layer 2
# Layer 3
topology = [2 * (shape**2), 100, 100]
# batch_size = 10
first_progress_logger = rbm_logger.ProgressLogger(img_shape=(shape * 2,
shape))
rest_progress_logger = rbm_logger.ProgressLogger()
first_rbm_config = rbm_config.RBMConfig(
train_params=base_tr, progress_logger=first_progress_logger)
first_rbm_config.v_unit = rbm_units.GaussianVisibleUnit
rest_rbm_config = rbm_config.RBMConfig(
train_params=rest_tr, progress_logger=rest_progress_logger)
rbm_configs = [first_rbm_config, rest_rbm_config, rest_rbm_config]
config = DBN.DBNConfig(topology=topology,
training_parameters=base_tr,
rbm_configs=rbm_configs,
data_manager=data_manager)
# construct the Deep Belief Network
dbn = DBN.DBN(config)
# Train DBN on concatenated images
train_tX = theano.function([], T.concatenate([train_x, train_x_ass],
axis=1))()
train_X = theano.shared(train_tX)
test_tX = theano.function([], T.concatenate([test_x, test_x_ass],
axis=1))()
test_X = theano.shared(test_tX)
test_tX2 = theano.function([],
T.concatenate(
[test_x, T.zeros_like(test_x)], axis=1))()
test_X2 = theano.shared(test_tX2)
origs = []
recons = []
recons2 = []
# Train DBN
dbn.pretrain(train_X,
cache=[True, True, False],
train_further=[True, True, True])
recon = dbn.reconstruct(train_X,
k=1,
plot_n=20,
img_name='stackedRBM_train_recon_{}_{}'.format(
topology, 0))
train_x_ass_recon = recon[:, shape**2:]
recon = dbn.reconstruct(test_X,
k=1,
plot_n=20,
img_name='stackedRBM_test_recon_{}_{}'.format(
topology, 0))
test_x_ass_recon = recon[:, shape**2:]
recon = dbn.reconstruct(test_X2,
k=2,
plot_n=20,
img_name='stackedRBM_test_zero_recon_{}_{}'.format(
topology, 0))
test_x_ass_recon2 = recon[:, shape**2:]
clf_recon = SimpleClassifier('logistic', train_x, train_y)
score_orig = clf_recon.get_score(test_x_ass_recon, test_y_ass.eval())
clf_recon.retrain(train_x_ass_recon, train_y_ass.eval())
score_recon = clf_recon.get_score(test_x_ass_recon, test_y_ass.eval())
score_recon2 = clf_recon.get_score(test_x_ass_recon2, test_y_ass.eval())
print 'classification rate: {}, {}, {}'.format(score_orig, score_recon,
score_recon2)
origs.append(score_orig)
recons.append(score_recon)
recons2.append(score_recon2)
def KanadeAssociativeDBN(cache=False):
print "Testing Associative RBM which tries to learn the following mapping: " \
"ID"
# "{anger, saddness, disgust} -> {sadness}, {contempt, happy, surprise} -> {happy}"
# project set-up
data_manager = store.StorageManager('Kanade/AssociativeDBNTest', log=True)
shape = 25
dataset_name = 'sharp_equi{}_{}'.format(shape, shape)
preprocessing = {'scale': True}
# Load kanade database
mapping = None
# mapping = {'anger': 'sadness',
# 'contempt': 'happy',
# 'disgust': 'sadness',
# 'fear': 'sadness',
# 'happy': 'happy',
# 'sadness': 'sadness',
# 'surprise': 'happy'}
dataset = loader.load_kanade(n=100,
pre=preprocessing,
set_name=dataset_name)
mapped_dataset = loader.load_kanade(
n=100,
# emotions=['sadness', 'happy'],
pre=preprocessing,
set_name=dataset_name) # Target Image
train, valid, test = dataset
train_x, train_y = train
test_x, test_y = test
# Sample associated image
train_x_ass, train_y_ass = loader.sample_image(train_y,
mapping=mapping,
pre=preprocessing,
set_name=dataset_name)
test_x_ass, test_y_ass = loader.sample_image(test_y,
mapping=mapping,
pre=preprocessing,
set_name=dataset_name)
# initialise AssociativeDBN
config = associative_dbn.DefaultADBNConfig()
# Gaussian Input Layer
bottom_tr = rbm_config.TrainParam(learning_rate=0.0001,
momentum_type=rbm_config.NESTEROV,
momentum=0.9,
weight_decay=0.0001,
epochs=20,
batch_size=10)
h_n = 150
bottom_logger = rbm_logger.ProgressLogger(img_shape=(shape, shape))
bottom_rbm = rbm_config.RBMConfig(v_unit=rbm_units.GaussianVisibleUnit,
v_n=shape**2,
h_n=h_n,
progress_logger=bottom_logger,
train_params=bottom_tr)
config.left_dbn.rbm_configs[0] = bottom_rbm
config.right_dbn.rbm_configs[0] = bottom_rbm
config.left_dbn.topology = [shape**2, h_n]
config.right_dbn.topology = [shape**2, h_n]
config.top_rbm.train_params.epochs = 20
config.top_rbm.train_params.batch_size = 10
config.n_association = 1000
config.reuse_dbn = True
adbn = associative_dbn.AssociativeDBN(config=config,
data_manager=data_manager)
# Plot sample
loader.save_faces(
train_x.get_value(borrow=True)[1:50],
tile=(10, 10),
img_name='n_orig.png',
)
loader.save_faces(train_x_ass.get_value(borrow=True)[1:50],
tile=(10, 10),
img_name='n_ass.png')
# Train classifier to be used for classifying reconstruction associated image layer
clf_orig = SimpleClassifier('knn', mapped_dataset[0][0],
mapped_dataset[0][1])
# Test DBN Performance
for i in xrange(0, 5):
# Train DBN - learn joint distribution
cache_left = [True]
cache_right = [True]
cache_top = False
cache = [cache_left, cache_right, cache_top]
adbn.train(train_x, train_x_ass, cache=cache)
print "... trained associative DBN"
# Reconstruct images
test_x_recon = adbn.recall(test_x, associate_steps=500, recall_steps=0)
print "... reconstructed images"
# Classify the reconstructions
# 1. Train classifier on original images
score_orig = clf_orig.get_score(test_x_recon, test_y_ass.eval())
# 2. Train classifier on reconstructed images - reconstruction obtained by right dbn
right_dbn = adbn.dbn_right
mapped_train_recon = right_dbn.reconstruct(
mapped_dataset[0][0],
k=1,
plot_n=100,
plot_every=1,
img_name='right_dbn_reconstruction')
clf_recon = SimpleClassifier('knn', mapped_train_recon,
mapped_dataset[0][1].eval())
score_retrain = clf_recon.get_score(test_x_recon, test_y_ass.eval())
out_msg = '{} (orig, retrain):{},{}'.format(adbn, score_orig,
score_retrain)
print out_msg
