def base(train_dp, valid_dp, logger, learning_rate):
# learning_rate = 0.01
rng = numpy.random.RandomState([2016,02,26])
max_epochs = 1000
cost = CECost()
stats = list()
test_dp = deepcopy(valid_dp)
train_dp.reset()
valid_dp.reset()
test_dp.reset()
# NETWORK TOPOLOGY:
model = MLP(cost=cost)
model.add_layer(Relu(idim=125, odim=125, irange=1.6, rng=rng))
model.add_layer(Softmax(idim=125, odim=19, rng=rng))
# define the optimiser, here stochasitc gradient descent
# with fixed learning rate and max_epochs
lr_scheduler = LearningRateFixed(
learning_rate=learning_rate, max_epochs=max_epochs)
optimiser = SGDOptimiser(lr_scheduler=lr_scheduler)
logger.info('Training started...')
tr_stats_b, valid_stats_b = optimiser.train(model, train_dp, valid_dp)
logger.info('Testing the model on test set:')
tst_cost, tst_accuracy = optimiser.validate(model, test_dp)
logger.info('ACL test set accuracy is %.2f %%, cost (%s) is %.3f' %
(tst_accuracy*100., cost.get_name(), tst_cost))
print "date: " + str(dt)
train_dp.reset()
test_dp.reset()
valid_dp.reset()
rng = numpy.random.RandomState([dt.year, dt.month, dt.day])
# define the model structure, here just one linear layer
# and mean square error cost
cost = CECost()
model = MLP(cost=cost)
model.add_layer(ConvRelu_Opt(1, 1, rng=rng, stride=(1, 1)))
model.add_layer(Sigmoid(idim=122, odim=122, rng=rng))
model.add_layer(Softmax(idim=122, odim=19, rng=rng))
#one can stack more layers here
# print map(lambda x: (x.idim, x.odim), model.layers)
lr_scheduler = LearningRateFixed(learning_rate=0.01, max_epochs=500)
optimiser = SGDOptimiser(lr_scheduler=lr_scheduler)
tr_stats, valid_stats = optimiser.train(model, train_dp, valid_dp)
tst_cost, tst_accuracy = optimiser.validate(model, test_dp)
seeds.append((tr_stats, valid_stats, (tst_cost, tst_accuracy)))
end = time.time()
print "scipy.correlate time: " + str(end - start)
with open('seeds_conv_fft_feat.pkl', 'wb') as f:
p.dump(seeds, f)
learning_rate = 0.07
max_epochs = 30
cost = CECost()
stats = list()
test_dp = deepcopy(valid_dp)
train_dp.reset()
valid_dp.reset()
test_dp.reset()
#define the model
model = MLP(cost=cost)
#model.add_layer(ComplexLinear(idim=125, odim=125, irange=1.6, rng=rng))
#model.add_layer(Sigmoid(idim=2*125, odim=125, irange=1.6, rng=rng))
model.add_layer(Sigmoid(idim=125, odim=125, irange=1.6, rng=rng))
model.add_layer(Softmax(idim=125, odim=19, rng=rng))
# define the optimiser, here stochasitc gradient descent
# with fixed learning rate and max_epochs
lr_scheduler = LearningRateFixed(learning_rate=learning_rate, max_epochs=max_epochs)
optimiser = SGDOptimiser(lr_scheduler=lr_scheduler)
logger.info('Training started...')
tr_stats, valid_stats = optimiser.train(model, train_dp, valid_dp)
logger.info('Testing the model on test set:')
tst_cost, tst_accuracy = optimiser.validate(model,test_dp )
logger.info('MNIST test set accuracy is %.2f %%, cost (%s) is %.3f'%(tst_accuracy*100., cost.get_name(), tst_cost))
optimiser = SGDOptimiser(lr_scheduler=lr_scheduler)
logger.info('Initialising data providers...')
train_dp = MNISTDataProvider(dset='train',
batch_size=250,
max_num_batches=-10,
randomize=True,
conv_reshape=True)
valid_dp = MNISTDataProvider(dset='valid',
batch_size=250,
max_num_batches=-10,
randomize=False,
conv_reshape=True)
logger.info('Training started...')
tsk8_2_tr_stats, tsk8_2_valid_stats = optimiser.train(tsk8_2_model, train_dp,
valid_dp)
logger.info('Testing the model on test set:')
test_dp = MNISTDataProvider(dset='eval',
batch_size=200,
max_num_batches=-10,
randomize=False,
conv_reshape=True)
tsk8_2_cost, tsk8_2_accuracy = optimiser.validate(tsk8_2_model, test_dp)
logger.info('MNIST test set accuracy is %.2f %% (cost is %.3f)' %
(tsk8_2_accuracy * 100., tsk8_2_cost))
#saving for future use
with open('tsk8_2_model.pkl', 'wb') as f:
cPickle.dump(tsk8_2_model, f)