NNetUtil Namespace Reference

Functions
double	unitStep (double x)
double	sigmoid (double x)
double	squareError (const std::vector< double > &v1, const std::vector< double > &v2)

Detailed Description

Utility functions for neural networks

Function Documentation

double NNetUtil::sigmoid

(

double

x

)

Sigmoid function

y(x) = 1 / (1 + exp(-x))

Parameters

x	input to sigmoid

Returns

output of sigmoid

double NNetUtil::squareError	(	const std::vector< double > &	v1,
		const std::vector< double > &	v2
	)

Calculate the square error of two vectors Scaled with (1/2) to normalize derivative

E(v1, v2) = (1/2) |v2 - v1|^2

Parameters

v1	first vector
v2	second vector

Returns

the square error of the two vectors

double NNetUtil::unitStep

(

double

x

)

Unit step function

y(x) = 0 if x < 0
y(x) = 1 if x >= 0

Parameters

x	input to step function

Returns

step function of input