mindspore.nn.SoftShrink

View Source On Gitee
class mindspore.nn.SoftShrink(lambd=0.5)[source]

Applies the SoftShrink function element-wise.

\[\begin{split}\text{SoftShrink}(x) = \begin{cases} x - \lambda, & \text{ if } x > \lambda \\ x + \lambda, & \text{ if } x < -\lambda \\ 0, & \text{ otherwise } \end{cases}\end{split}\]

SoftShrink Activation Function Graph:

../../_images/Softshrink.png
Parameters

lambd (number, optional) – The threshold \(\lambda\) defined by the Soft Shrink formula. It should be greater than or equal to 0, default: 0.5 .

Inputs:

  • input (Tensor) - The input of Soft Shrink. Supported dtypes:

    • Ascend: float16, float32, bfloat16.

    • CPU/GPU: float16, float32.

Outputs:

Tensor, the same shape and data type as the input.

Raises

  • TypeError – If lambd is not a float, int or bool.

  • TypeError – If input is not a tensor.

  • TypeError – If dtype of input is not float16, float32 or bfloat16.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore
>>> from mindspore import Tensor, nn
>>> import numpy as np
>>> input = Tensor(np.array([[ 0.5297,  0.7871,  1.1754], [ 0.7836,  0.6218, -1.1542]]), mindspore.float16)
>>> softshrink = nn.SoftShrink()
>>> output = softshrink(input)
>>> print(output)
[[ 0.02979  0.287    0.676  ]
 [ 0.2837   0.1216  -0.6543 ]]