activationfunction

Documentation

Overview

Package activationfunction is for isolating the concept of an activation function into its own area. Activation functions are used in neural networks to provide stability to their inputs and outputs through non-linearization. What that means is that a neuron which is loaded up with tons of values in its weighted sum will be sent through an activation function which sets a bound on it to prevent this large value from influencing the rest of the network.

Index

• func ErrNotFound(name Name) error
• type ActivationFunction
  • func GetFunction(name Name) (ActivationFunction, error)
  • func MustGetFunction(name Name) ActivationFunction
• type Name

Functions

func ErrNotFound

func ErrNotFound(name Name) error

Types

type ActivationFunction

type ActivationFunction func(x float64) float64

func GetFunction

func GetFunction(name Name) (ActivationFunction, error)

func MustGetFunction

func MustGetFunction(name Name) ActivationFunction

MustGetFunction calls GetFunction but panics if an error is encountered.

type Name

type Name string

const (
	NameNoop    Name = "noop"
	NameSigmoid Name = "sigmoid"
	NameTanh    Name = "tanh"
	NameRelu    Name = "relu"
	NameLinear  Name = "linear"
)