def train_dbn(dbn_weights, training_data, learning_rate=0.1, max_epochs=1):
# train layer by layer
# fully train one layer, then train the next, until all are trained
for idx, weight in enumerate(dbn_weights):
if idx == 0: # first layer training
rbm.train(weight, training_data, learning_rate, max_epochs)
continue
else:
training_data = rbm.get_states(rbm.construct(dbn_weights[idx - 1], training_data))
rbm.train(weight, training_data, learning_rate, max_epochs)
return dbn_weights
def train_dbn(dbn_weights, training_data, learning_rate = .1, max_epochs = 1):
#train layer by layer
#fully train one layer, then train the next, until all are trained
for idx, weight in enumerate(dbn_weights):
if idx == 0: #first layer training
rbm.train(weight, training_data, learning_rate, max_epochs)
continue
else:
training_data = rbm.get_states(rbm.construct(dbn_weights[idx-1], training_data))
rbm.train(weight, training_data,learning_rate, max_epochs)
return dbn_weights
def construct(weights, vis_states):
for weight in weights:
vis_states = rbm.construct(weight, vis_states)
return vis_states
def construct(weights, vis_states):
for weight in weights:
vis_states = rbm.construct(weight, vis_states)
return vis_states
