Nonlinear Activation Functions

This library provides several nonlinear functions commonly used as activation functions in neural networks.

Interface

To use this library, prefix the types and functions below with Nonlinear..

Methods

• val sigmoid : real -> real
• val dsigmoid : real -> real
• val tanh : real -> real
• val dtanh : real -> real
• val sigmoidf : real -> real
• val sigmoidh : real -> real
• val sigmoidg : real -> real
• val softmax : real array * int -> real
• val exp : real -> real
• val log : real -> real

Method Overview

• sigmoid x
  • Sigmoid function. Returns 1.0 / (1.0 + (e^(-x))).
• dsigmoid x
  • Derivative of sigmoid.
• tanh (x)
  • The hyperbolic tangent.
• dtanh x.
  • Derivative of tanh.
• sigmoidf x
  • Alias for the sigmoid function.
• sigmoidh x
  • Returns 2.0 / (1.0 + (e^(-x))) - 1.0.
• sigmoidg x
  • Returns 4.0 / (1.0 + (e^(-x))) - 2.0.
• softmax (arr, n)
  • The softmax function.
• exp x
  • Returns the base of the natural logarithm raised to the xth power.
• log x
  • The natural logarithm.