fml::dimops Namespace Reference

Row/column operations. More...

Enumerations
enum	sweep_op { SWEEP_ADD, SWEEP_SUB, SWEEP_MUL, SWEEP_DIV }

Functions
template<typename REAL >
void	rowsums (const cpumat< REAL > &x, cpuvec< REAL > &s)
	Compute the row sums. More...
template<typename REAL >
void	rowmeans (const cpumat< REAL > &x, cpuvec< REAL > &s)
	Compute the row means. More...
template<typename REAL >
void	colsums (const cpumat< REAL > &x, cpuvec< REAL > &s)
	Compute the column sums. More...
template<typename REAL >
void	colmeans (const cpumat< REAL > &x, cpuvec< REAL > &s)
	Compute the column means. More...
template<typename REAL >
void	scale (const bool rm_mean, const bool rm_sd, cpumat< REAL > &x)
	Remove the mean and/or the sd from a matrix. More...
template<typename REAL >
void	rowsums (const gpumat< REAL > &x, gpuvec< REAL > &s)
	Compute the row sums. More...
template<typename REAL >
void	rowmeans (const gpumat< REAL > &x, gpuvec< REAL > &s)
	Compute the row means. More...
template<typename REAL >
void	colsums (const gpumat< REAL > &x, gpuvec< REAL > &s)
	Compute the column sums. More...
template<typename REAL >
void	colmeans (const gpumat< REAL > &x, gpuvec< REAL > &s)
	Compute the column means. More...
template<typename REAL >
void	scale (const bool rm_mean, const bool rm_sd, gpumat< REAL > &x)
	Remove the mean and/or the sd from a matrix. More...
template<typename REAL >
void	rowsums (const mpimat< REAL > &x, cpuvec< REAL > &s)
	Compute the row sums. More...
template<typename REAL >
void	rowmeans (const mpimat< REAL > &x, cpuvec< REAL > &s)
	Compute the row means. More...
template<typename REAL >
void	colsums (const mpimat< REAL > &x, cpuvec< REAL > &s)
	Compute the column sums. More...
template<typename REAL >
void	colmeans (const mpimat< REAL > &x, cpuvec< REAL > &s)
	Compute the column means. More...
template<typename REAL >
void	scale (const bool rm_mean, const bool rm_sd, mpimat< REAL > &x)
	Remove the mean and/or the sd from a matrix. More...

Detailed Description

Row/column operations.

Enumeration Type Documentation

sweep_op

enum fml::dimops::sweep_op

Enumerator
SWEEP_ADD	Arithmetic operations for dimops::rowsweep() and dimops::colsweep() functions. These operate by either row or column-wise taking a matrix and applying an appropriately sized vector to the entries of that matrix by addition, subtraction, multiplication, or division.

Function Documentation

colmeans() [1/3]

template<typename REAL >

void fml::dimops::colmeans	(	const cpumat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the column means.

Parameters

[in]	x	Input data.
[out]	s	Column means.

Template Parameters

REAL	should be 'float' or 'double'.

colmeans() [2/3]

template<typename REAL >

void fml::dimops::colmeans	(	const gpumat< REAL > &	x,
		gpuvec< REAL > &	s
	)

Compute the column means.

Parameters

[in]	x	Input data.
[out]	s	Column means.

Implementation Details
Uses linalg::matmult() on a vector of ones.

Memory Allocations
If the output is inappropriately sized, it will automatically be re-allocated. Additionally, some temporary work storage is needed.

Exceptions
If a (re-)allocation is triggered and fails, a bad_alloc exception will be thrown.

Template Parameters

REAL	should be 'float' or 'double'.

colmeans() [3/3]

template<typename REAL >

void fml::dimops::colmeans	(	const mpimat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the column means.

Parameters

[in]	x	Input data.
[out]	s	Column means.

Template Parameters

REAL	should be 'float' or 'double'.

colsums() [1/3]

template<typename REAL >

void fml::dimops::colsums	(	const cpumat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the column sums.

Parameters

[in]	x	Input data.
[out]	s	Column sums.

Template Parameters

REAL	should be 'float' or 'double'.

colsums() [2/3]

template<typename REAL >

void fml::dimops::colsums	(	const gpumat< REAL > &	x,
		gpuvec< REAL > &	s
	)

Compute the column sums.

Parameters

[in]	x	Input data.
[out]	s	Column sums.

Implementation Details
Uses linalg::matmult() on a vector of ones.

Memory Allocations
If the output is inappropriately sized, it will automatically be re-allocated. Additionally, some temporary work storage is needed.

Exceptions
If a (re-)allocation is triggered and fails, a bad_alloc exception will be thrown.

Template Parameters

REAL	should be 'float' or 'double'.

colsums() [3/3]

template<typename REAL >

void fml::dimops::colsums	(	const mpimat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the column sums.

Parameters

[in]	x	Input data.
[out]	s	Column sums.

Template Parameters

REAL	should be 'float' or 'double'.

rowmeans() [1/3]

template<typename REAL >

void fml::dimops::rowmeans	(	const cpumat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the row means.

Parameters

[in]	x	Input data.
[out]	s	Row means.

Template Parameters

REAL	should be 'float' or 'double'.

rowmeans() [2/3]

template<typename REAL >

void fml::dimops::rowmeans	(	const gpumat< REAL > &	x,
		gpuvec< REAL > &	s
	)

Compute the row means.

Parameters

[in]	x	Input data.
[out]	s	Row means.

Implementation Details
Uses linalg::matmult() on a vector of ones.

Memory Allocations
If the output is inappropriately sized, it will automatically be re-allocated. Additionally, some temporary work storage is needed.

Exceptions
If a (re-)allocation is triggered and fails, a bad_alloc exception will be thrown.

Template Parameters

REAL	should be 'float' or 'double'.

rowmeans() [3/3]

template<typename REAL >

void fml::dimops::rowmeans	(	const mpimat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the row means.

Parameters

[in]	x	Input data.
[out]	s	Row means.

Template Parameters

REAL	should be 'float' or 'double'.

rowsums() [1/3]

template<typename REAL >

void fml::dimops::rowsums	(	const cpumat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the row sums.

Parameters

[in]	x	Input data.
[out]	s	Row sums.

Template Parameters

REAL	should be 'float' or 'double'.

rowsums() [2/3]

template<typename REAL >

void fml::dimops::rowsums	(	const gpumat< REAL > &	x,
		gpuvec< REAL > &	s
	)

Compute the row sums.

Parameters

[in]	x	Input data.
[out]	s	Row sums.

Implementation Details
Uses linalg::matmult() on a vector of ones.

Memory Allocations
If the output is inappropriately sized, it will automatically be re-allocated. Additionally, some temporary work storage is needed.

Exceptions
If a (re-)allocation is triggered and fails, a bad_alloc exception will be thrown.

Template Parameters

REAL	should be 'float' or 'double'.

rowsums() [3/3]

template<typename REAL >

void fml::dimops::rowsums	(	const mpimat< REAL > &	x,
		cpuvec< REAL > &	s
	)

Compute the row sums.

Parameters

[in]	x	Input data.
[out]	s	Row sums.

Template Parameters

REAL	should be 'float' or 'double'.

scale() [1/3]

template<typename REAL >

void fml::dimops::scale	(	const bool	rm_mean,
		const bool	rm_sd,
		cpumat< REAL > &	x
	)

Remove the mean and/or the sd from a matrix.

Parameters

[in]	rm_mean	Remove the column means?
[in]	rm_sd	Remove the column sds?
[in,out]	x	Data to center/scale.

Template Parameters

REAL	should be 'float' or 'double'.

scale() [2/3]

template<typename REAL >

void fml::dimops::scale	(	const bool	rm_mean,
		const bool	rm_sd,
		gpumat< REAL > &	x
	)

Remove the mean and/or the sd from a matrix.

Parameters

[in]	rm_mean	Remove the column means?
[in]	rm_sd	Remove the column sds?
[in,out]	x	Data to center/scale.

Memory Allocations
Some temporary work storage is needed to store vectors of means and/or standard deviations, depending on what is requested.

Exceptions
If an allocation fails, a bad_alloc exception will be thrown.

Template Parameters

REAL	should be 'float' or 'double'.

scale() [3/3]

template<typename REAL >

void fml::dimops::scale	(	const bool	rm_mean,
		const bool	rm_sd,
		mpimat< REAL > &	x
	)

Remove the mean and/or the sd from a matrix.

Parameters

[in]	rm_mean	Remove the column means?
[in]	rm_sd	Remove the column sds?
[in,out]	x	Data to center/scale.

Template Parameters

REAL	should be 'float' or 'double'.