Numerix v0.6.0 Numerix.Activations View Source

Activation functions for neural networks.

Link to this section Summary

Functions

elu(x, alpha \\ 1.0)

Computes the exponential linear unit of a tensor.

leaky_relu(x, alpha)

Computes the leaky rectified linear unit of a tensor.

relu(x)

Computes the rectified linear unit of a tensor.

selu(x)

Computes the scaled exponential linear unit of a tensor.

Computes the element-wise sigmoid of a tensor.

Computes the softmax of a tensor.

Computes the softplus of a tensor.

Computes the softsign of a tensor.

tanh(x)

Computes the element-wise hyperbolic tangent of a tensor.

Link to this section Functions

elu(x, alpha \\ 1.0)

elu(%Numerix.Tensor{dims: term(), items: term(), shape: term()}, number()) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the exponential linear unit of a tensor.

leaky_relu(x, alpha)

leaky_relu(
  %Numerix.Tensor{dims: term(), items: term(), shape: term()},
  number()
) :: %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the leaky rectified linear unit of a tensor.

relu(x)

relu(%Numerix.Tensor{dims: term(), items: term(), shape: term()}) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the rectified linear unit of a tensor.

selu(x)

selu(%Numerix.Tensor{dims: term(), items: term(), shape: term()}) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the scaled exponential linear unit of a tensor.

sigmoid(x)

sigmoid(%Numerix.Tensor{dims: term(), items: term(), shape: term()}) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the element-wise sigmoid of a tensor.

softmax(x)

softmax(%Numerix.Tensor{dims: term(), items: term(), shape: term()}) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the softmax of a tensor.

softplus(x)

softplus(%Numerix.Tensor{dims: term(), items: term(), shape: term()}) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the softplus of a tensor.

softsign(x)

softsign(%Numerix.Tensor{dims: term(), items: term(), shape: term()}) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the softsign of a tensor.

tanh(x)

tanh(%Numerix.Tensor{dims: term(), items: term(), shape: term()}) ::
  %Numerix.Tensor{dims: term(), items: term(), shape: term()}

Computes the element-wise hyperbolic tangent of a tensor.