Skip to content

Tanh — Hyperbolic Tangent Activation

Squashes inputs to (-1, 1) with zero-centered output. Useful when subsequent layers benefit from positive and negative responses of similar scale.


Overview

Hyperbolic tangent is smooth, zero-centered, and bounded. It avoids the positive-bias of ReLU and can help with gradient symmetry in deep stacks.


Key Features

  • Zero-centered output in (-1, 1)
  • Smooth, fully differentiable
  • Avoids positive-only bias of ReLU

Parameters

Parameter Type Default Description
name str \| None None Layer name
verbose bool False Debug logging

Methods

forward(x, training=True)

Return the activation applied to input.

  • x: Tensor or array-like — Input values of any shape
  • Returns: Tensor with same shape as x

Example Usage

import numpy as np
from machinegnostics.magnet import Tanh

activation = Tanh()
result = activation(np.array([[0.0]]))
print(result.data.tolist())  # [[0.0]]

Author: Nirmal Parmar