linear regression

Collaboration diagram for linear regression:

Data Structures
struct	a_regress_linear
	instance structure for linear regression More...

Typedefs
typedef struct a_regress_linear	a_regress_linear

Functions
void	a_regress_linear_init (a_regress_linear *ctx, a_real *coef_p, a_size coef_n, a_real bias)
	initialize for linear regression
a_real	a_regress_linear_eval (a_regress_linear const *ctx, a_real const *val)
	compute predicted value for linear regression
void	a_regress_linear_err (a_regress_linear const *ctx, a_size n, a_real const *x, a_real const *y, a_real *err)
	compute residuals for linear regression
void	a_regress_linear_pdm (a_regress_linear const *ctx, a_size n, a_real const *x, a_real *pdm, a_real y_mean)
	compute prediction deviation from mean for linear regression
void	a_regress_linear_gd (a_regress_linear *ctx, a_real const *input, a_real error, a_real alpha)
	gradient descent for linear regression
void	a_regress_linear_sgd (a_regress_linear *ctx, a_size n, a_real const *x, a_real const *y, a_real alpha)
	stochastic gradient descent for linear regression
void	a_regress_linear_bgd (a_regress_linear *ctx, a_size n, a_real const *x, a_real const *err, a_real alpha)
	batch gradient descent for linear regression
a_real	a_regress_linear_mgd (a_regress_linear *ctx, a_size n, a_real const *x, a_real const *y, a_real *err, a_real delta, a_real lrmax, a_real lrmin, a_size lrtim, a_size epoch, a_size batch)
	mini-batch gradient descent for linear regression
void	a_regress_linear_zero (a_regress_linear *ctx)
	zeroing for linear regression

Detailed Description

Function Documentation

a_regress_linear_bgd()

void a_regress_linear_bgd	(	a_regress_linear *	ctx,
		a_size	n,
		a_real const *	x,
		a_real const *	err,
		a_real	alpha )

batch gradient descent for linear regression

Parameters

[in,out]	ctx	points to an instance of linear regression
[in]	n	number of samples, x[n*coef_n], err[n]
[in]	x	predictor data, specified as a numeric matrix
[in]	err	residuals, specified as a numeric vector
[in]	alpha	learning rate for gradient descent

a_regress_linear_err()

void a_regress_linear_err	(	a_regress_linear const *	ctx,
		a_size	n,
		a_real const *	x,
		a_real const *	y,
		a_real *	err )

compute residuals for linear regression

Parameters

[in]	ctx	points to an instance of linear regression
[in]	n	number of samples, x[n*coef_n], y[n], err[n]
[in]	x	predictor data, specified as a numeric matrix
[in]	y	response data, specified as a numeric vector
[in,out]	err	residuals, specified as a numeric vector

a_regress_linear_eval()

a_real a_regress_linear_eval	(	a_regress_linear const *	ctx,
		a_real const *	val )

compute predicted value for linear regression

Parameters

[in]	ctx	points to an instance of linear regression
[in]	val	independent variables

Returns

predicted value

a_regress_linear_gd()

void a_regress_linear_gd	(	a_regress_linear *	ctx,
		a_real const *	input,
		a_real	error,
		a_real	alpha )

gradient descent for linear regression

Parameters

[in,out]	ctx	points to an instance of linear regression
[in]	input	predictor data, specified as a numeric vector
[in]	error	residual, specified as a numeric scalar
[in]	alpha	learning rate for gradient descent

a_regress_linear_init()

void a_regress_linear_init	(	a_regress_linear *	ctx,
		a_real *	coef_p,
		a_size	coef_n,
		a_real	bias )

initialize for linear regression

Parameters

[in,out]	ctx	points to an instance of linear regression
[in]	coef_p	points to regression coefficients
[in]	coef_n	number of regression coefficients
[in]	bias	intercept

a_regress_linear_mgd()

a_real a_regress_linear_mgd	(	a_regress_linear *	ctx,
		a_size	n,
		a_real const *	x,
		a_real const *	y,
		a_real *	err,
		a_real	delta,
		a_real	lrmax,
		a_real	lrmin,
		a_size	lrtim,
		a_size	epoch,
		a_size	batch )

mini-batch gradient descent for linear regression

Parameters

[in,out]	ctx	points to an instance of linear regression
[in]	n	number of samples, x[n*coef_n], y[n], err[n]
[in]	x	predictor data, specified as a numeric matrix
[in]	y	response data, specified as a numeric vector
[in,out]	err	residuals, specified as a numeric vector
[in]	delta	threshold for gradient descent value
[in]	lrmax	maximum learning rate of iterations
[in]	lrmin	minimum learning rate of iterations
[in]	lrtim	total number of learning rate steps
[in]	epoch	maximum number of epochs
[in]	batch	batch size of data

Returns

change in loss function

a_regress_linear_pdm()

void a_regress_linear_pdm	(	a_regress_linear const *	ctx,
		a_size	n,
		a_real const *	x,
		a_real *	pdm,
		a_real	y_mean )

compute prediction deviation from mean for linear regression

Parameters

[in]	ctx	points to an instance of linear regression
[in]	n	number of samples, x[n*coef_n], pdm[n]
[in]	x	predictor data, specified as a numeric matrix
[in,out]	pdm	deviation, specified as a numeric vector
[in]	y_mean	mean of response data

a_regress_linear_sgd()

void a_regress_linear_sgd	(	a_regress_linear *	ctx,
		a_size	n,
		a_real const *	x,
		a_real const *	y,
		a_real	alpha )

stochastic gradient descent for linear regression

Parameters

[in,out]	ctx	points to an instance of linear regression
[in]	n	number of samples, x[n*coef_n], y[n]
[in]	x	predictor data, specified as a numeric matrix
[in]	y	response data, specified as a numeric vector
[in]	alpha	learning rate for gradient descent

a_regress_linear_zero()

void a_regress_linear_zero

(

a_regress_linear *

ctx

)

zeroing for linear regression

Parameters

[in,out]

ctx

points to an instance of linear regression