mshadow::Random< gpu, DType > Class Template Reference

GPU random number generator. More...

#include <random.h>

Collaboration diagram for mshadow::Random< gpu, DType >:

Public Member Functions
	Random (int seed)
	constructor of random engine More...
	~Random (void) MSHADOW_THROW_EXCEPTION
void	set_stream (Stream< gpu > *stream)
	set the stream of computation More...
void	Seed (int seed)
	seed random number generator using this seed More...
void	GetRandInt (const Tensor< gpu, 1, unsigned > &dst)
	get a set of random integers More...
template<int dim>
void	SampleUniform (Tensor< gpu, dim, DType > *dst, DType a=0.0f, DType b=1.0f)
	generate data from uniform [a,b) More...
template<int dim>
void	SampleGaussian (Tensor< gpu, dim, DType > *dst, DType mu=0.0f, DType sigma=1.0f)
	generate data from standard gaussian More...
template<int dim>
expr::ReshapeExp< Tensor< gpu, 1, DType >, DType, dim, 1 >	gaussian (Shape< dim > shape, DType mu=0.0f, DType sigma=1.0f)
	return a temporal expression storing standard gaussian random variables the temporal tensor is only valid before next call of gaussian or uniform can be used as part of expression Caution: this means expression such as A = gaussian(s1) * gaussian(s2) will give invalid result, since second call of gaussian(s2) makes gaussian(s1) invalid A = gaussian(s1)*B+C; is correct; use one gaussian/uniform in each expression More...
template<int dim>
expr::ReshapeExp< Tensor< gpu, 1, DType >, DType, dim, 1 >	uniform (Shape< dim > shape)
	return a temporal expression storing standard uniform [0,1) the temporal tensor is only valid before next call of gaussian or uniform can be used as part of expression Caution: this means expression such as A = gaussian(s1) * gaussian(s2) will give invalid result, since second call of gaussian(s2) makes gaussian(s1) invalid A = gaussian(s1)*B+C; is correct; use one gaussian/uniform in each expression More...

Detailed Description

template<typename DType>
class mshadow::Random< gpu, DType >

GPU random number generator.

Constructor & Destructor Documentation

template<typename DType >

mshadow::Random< gpu, DType >::Random ( int  seed )

inlineexplicit

constructor of random engine

Parameters

seed	random number seed

template<typename DType >

mshadow::Random< gpu, DType >::~Random ( void  )

inline

Member Function Documentation

template<typename DType >

template<int dim>

expr::ReshapeExp<Tensor<gpu, 1, DType>, DType, dim, 1> mshadow::Random< gpu, DType >::gaussian ( Shape< dim >  shape, DType  mu = 0.0f, DType  sigma = 1.0f  )

inline

return a temporal expression storing standard gaussian random variables the temporal tensor is only valid before next call of gaussian or uniform can be used as part of expression Caution: this means expression such as A = gaussian(s1) * gaussian(s2) will give invalid result, since second call of gaussian(s2) makes gaussian(s1) invalid A = gaussian(s1)*B+C; is correct; use one gaussian/uniform in each expression

Parameters

shape	shape of the tensor
mu	mean
sigma	variance

Returns

a temporal expression storing standard gaussian random variables

Template Parameters

dim	dimension of tensor

template<typename DType >

void mshadow::Random< gpu, DType >::GetRandInt ( const Tensor< gpu, 1, unsigned > &  dst )

inline

get a set of random integers

template<typename DType >

template<int dim>

void mshadow::Random< gpu, DType >::SampleGaussian ( Tensor< gpu, dim, DType > *  dst, DType  mu = 0.0f, DType  sigma = 1.0f  )

inline

generate data from standard gaussian

Parameters

dst	destination
mu	mean variable
sigma	standard deviation

Template Parameters

dim	dimension of tensor

template<typename DType >

template<int dim>

void mshadow::Random< gpu, DType >::SampleUniform ( Tensor< gpu, dim, DType > *  dst, DType  a = 0.0f, DType  b = 1.0f  )

inline

generate data from uniform [a,b)

Parameters

dst	destination
a	lower bound of uniform
b	upper bound of uniform

Template Parameters

dim	dimension of tensor

template<typename DType >

void mshadow::Random< gpu, DType >::Seed ( int  seed )

inline

seed random number generator using this seed

Parameters

seed	seed of prng

template<typename DType >

void mshadow::Random< gpu, DType >::set_stream ( Stream< gpu > *  stream )

inline

set the stream of computation

Parameters

stream

computation stream

template<typename DType >

template<int dim>

expr::ReshapeExp<Tensor<gpu, 1, DType>, DType, dim, 1> mshadow::Random< gpu, DType >::uniform ( Shape< dim >  shape )

inline

return a temporal expression storing standard uniform [0,1) the temporal tensor is only valid before next call of gaussian or uniform can be used as part of expression Caution: this means expression such as A = gaussian(s1) * gaussian(s2) will give invalid result, since second call of gaussian(s2) makes gaussian(s1) invalid A = gaussian(s1)*B+C; is correct; use one gaussian/uniform in each expression

Parameters

shape

shape of the tensor

Returns

a temporal expression storing standard uniform [0,1)

Template Parameters

dim	dimension of tensor

---

The documentation for this class was generated from the following file:

• /work/mxnet/3rdparty/mshadow/mshadow/random.h