def __init__(self,
layers,
optimizer,
loss,
max_epochs=10,
batch_size=64,
metric='mse',
shuffle=False,
verbose=True):
self.verbose = verbose
self.shuffle = shuffle
self.optimizer = optimizer
self.loss = get_loss(loss)
# TODO: fix
if loss == 'categorical_crossentropy':
self.loss_grad = lambda actual, predicted: -(actual - predicted)
else:
self.loss_grad = elementwise_grad(self.loss, 1)
self.metric = get_metric(metric)
self.layers = layers
self.batch_size = batch_size
self.max_epochs = max_epochs
self._n_layers = 0
self.log_metric = True if loss != metric else False
self.metric_name = metric
self.bprop_entry = self._find_bprop_entry()
self.training = False
self._initialized = False
def __init__(self, layers, optimizer, loss, max_epochs=10, batch_size=64, metric='mse',
shuffle=False, verbose=True):
self.verbose = verbose
self.shuffle = shuffle
self.optimizer = optimizer
self.loss = get_loss(loss)
# TODO: fix
if loss == 'categorical_crossentropy':
self.loss_grad = lambda actual, predicted: -(actual - predicted)
else:
self.loss_grad = elementwise_grad(self.loss, 1)
self.metric = get_metric(metric)
self.layers = layers
self.batch_size = batch_size
self.max_epochs = max_epochs
self._n_layers = 0
self.log_metric = True if loss != metric else False
self.metric_name = metric
self.bprop_entry = self._find_bprop_entry()
self.training = False
self._initialized = False
def metric(name):
return validate_input(get_metric(name))