Activations¶
Paddle Supported Activations.
Each activation inherit BaseActivation, which has two attributes.
name: activation name in paddle config.
- support_hppl: True if supported by hppl. lstm layer can only use activations
supported by hppl (the name hppl will be revised later).
-
class
paddle.trainer_config_helpers.activations.BaseActivation(name, support_hppl)¶ A mark for activation class.
-
class
paddle.trainer_config_helpers.activations.TanhActivation¶ Tanh activation.
\[f(z)=tanh(z)=\frac{e^z-e^{-z}}{e^z+e^{-z}}\]
-
class
paddle.trainer_config_helpers.activations.SigmoidActivation¶ Sigmoid activation.
\[f(z) = \frac{1}{1+exp(-z)}\]
-
class
paddle.trainer_config_helpers.activations.SoftmaxActivation¶ Softmax activation for simple input
\[P(y=j|x) = \frac{e^{x_j}} {\sum^K_{k=1} e^{x_j} }\]
-
class
paddle.trainer_config_helpers.activations.SequenceSoftmaxActivation¶ Softmax activation for one sequence. The dimension of input feature must be 1 and a sequence.
result = softmax(for each_feature_vector[0] in input_feature) for i, each_time_step_output in enumerate(output): each_time_step_output = result[i]
-
class
paddle.trainer_config_helpers.activations.IdentityActivation¶ Identity Activation.
Just do nothing for output both forward/backward.
-
paddle.trainer_config_helpers.activations.LinearActivation¶ alias of
IdentityActivation
-
class
paddle.trainer_config_helpers.activations.ReluActivation¶ Relu activation.
forward. \(y = max(0, z)\)
derivative:
\[\begin{split}1 &\quad if z > 0 \\ 0 &\quad\mathrm{otherwize}\end{split}\]
-
class
paddle.trainer_config_helpers.activations.BReluActivation¶ BRelu Activation.
forward. \(y = min(24, max(0, z))\)
derivative:
\[\begin{split}1 &\quad if 0 < z < 24 \\ 0 &\quad \mathrm{otherwise}\end{split}\]
-
class
paddle.trainer_config_helpers.activations.SoftReluActivation¶ SoftRelu Activation.
-
class
paddle.trainer_config_helpers.activations.STanhActivation¶ Scaled Tanh Activation.
\[f(z) = 1.7159 * tanh(2/3*z)\]
-
class
paddle.trainer_config_helpers.activations.AbsActivation¶ Abs Activation.
Forward: \(f(z) = abs(z)\)
Derivative:
\[\begin{split}1 &\quad if \quad z > 0 \\ -1 &\quad if \quad z < 0 \\ 0 &\quad if \quad z = 0\end{split}\]
-
class
paddle.trainer_config_helpers.activations.SquareActivation¶ Square Activation.
\[f(z) = z^2.\]