ActivationFunctions

ActivationFunctions.relu(z)[source]
ActivationFunctions.clipped01lu(z)[source]

0 for x <= 0 x for 0 <= x <= 1 1 for 1 <= x

ActivationFunctions.clippedlu(z)[source]
-1 for x <= -1 x for -1 <= x <= 1 1 for 1 <= x
ActivationFunctions.elu(z)[source]
ActivationFunctions.identity(z)[source]
ActivationFunctions.softsign(z)[source]
ActivationFunctions.softsquare(z)[source]
ActivationFunctions.maxout(z)[source]
ActivationFunctions.softmax(z)[source]
ActivationFunctions.gauss(z)[source]
ActivationFunctions.cdf(z)[source]

Cumulative distribution function via erf (Error function)

ActivationFunctions.constant_one()[source]
ActivationFunctions.constant_zero()[source]
class ActivationFunctions.Round3(output_types_preference=None, name=None)[source]
c_code(node, name, x, z, sub)[source]
grad(inputs, gout)[source]
ActivationFunctions.hard_sigmoid(x)[source]
ActivationFunctions.binary_tanh(x)[source]
ActivationFunctions.binary_sigmoid(x)[source]
ActivationFunctions.strtoact(act)[source]
Parameters:act (str | list[str]) – activation function name, or multiple such as a list or separated by ”:”
Return type:theano.Op | list[theano.Op]
ActivationFunctions.strtoact_single_joined(act)[source]
Parameters:act (str | None) – activation function name, or multiple such as a list or separated by ”:”
Return type:theano.Op