The Complete SPRLIB & ANNLIB

fisher: - compute the optimal linear discriminant function for two classes

SYNOPSIS

int fisher

(Ka, Kb, Ma, Mb, c, n, W, b)

ARGUMENTS

`double`	`Ka`**	Covariance matrix for class A, a `matrix` of size `[1..n][1..n]`.
`double`	`Ka`**	Covariance matrix for class B, a `matrix` of size `[1..n][1..n]`.
`double`	*`Ma`**	Mean vector for class A, a `vector` of size `[1..n]`.
`double`	*`Mb`**	Mean vector for class B, a `vector` of size `[1..n]`.
`double`	`c`	The factor for apriori probabilities.
`int`	`n`	The dimensionality of the patterns.
`double`	*`W`**	The linear term, a `vector` of size `[1..n]`.
`double`	`b`	The constant component.

RETURNS

The function returns TRUE if an error occurred. If no error occurred the function returns FALSE and W and b contain valid values.

FUNCTION

This function computes the terms for the Fisher linear discriminant. The discrimimant is described by W*x + b. In order to calculate these parameters the covariances and means of both classes are needed. Furthermore the apriori probabilities are needed to compensate for inequalities. The parameter c is used for that since c equals P(b)/P(a) where P(a) and P(b) are the apriori probabilites for class A and B respectively.

NOTE

The matrices Ka ,Kb are in matrix format and the variables Ma, Mb, W and b are in vector format.

The Complete SPRLIB & ANNLIB

`fisher`

- compute the optimal linear discriminant function for two classes

SYNOPSIS

ARGUMENTS

RETURNS

FUNCTION

NOTE

SEE ALSO