OpenANN::DataStream Class Reference

Streams training data for online training. More...

#include <DataStream.h>

Public Member Functions
	DataStream (int cacheSize)
	Create a data stream. More...
	~DataStream ()
DataStream &	setLearner (Learner &learner)
	Add a learner. More...
DataStream &	setOptimizer (Optimizer &opt)
	Add an optimizer. More...
void	addSample (Eigen::VectorXd *x, Eigen::VectorXd *t=0)
	Add a sample of the data distribution. More...

Detailed Description

Streams training data for online training.

A DataStream combines a learner and an optimizer. It is able to cache training data and will start an optimization epoch if the cache is full. Training instances will be added incrementally with DataStream::addSample(). The optimization algorithm that will be passed with DataStream::setOptimizer() should work for online learning like e.g. MBSGD.

Constructor & Destructor Documentation

OpenANN::DataStream::DataStream

(

int

cacheSize)

Create a data stream.

Parameters

cacheSize

size of the internal data cache

OpenANN::DataStream::~DataStream

(

)

Member Function Documentation

void OpenANN::DataStream::addSample	(	Eigen::VectorXd *	x,
		Eigen::VectorXd *	t = 0
	)

Add a sample of the data distribution.

This will start an optimization epoch if the cache is full.

Parameters

x	input
t	target (can be 0 for unsupervised learning)

DataStream& OpenANN::DataStream::setLearner

(

Learner &

learner)

Add a learner.

Parameters

learner

model

Returns

this for chaining

DataStream& OpenANN::DataStream::setOptimizer

(

Optimizer &

opt)

Add an optimizer.

Parameters

opt	online optimization algorithm

Returns

this for chaining

---

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

• OpenANN/io/DataStream.h