com.mapr.stats.bandit

Class BayesianBandit

java.lang.Object

com.mapr.stats.bandit.BayesianBandit

All Implemented Interfaces:

Iterable<AbstractBayesianDistribution>

Direct Known Subclasses:

BetaBayesModel, EpsilonGreedy, GammaNormalBayesModel

public abstract class BayesianBandit
extends Object
implements Iterable<AbstractBayesianDistribution>

Implements the common characteristics of the Bayesian Bandit. All that is missing is the specific distribution that is used to learn and to sample the mean returns.

Typically, an implementation only has to call addModelDistribution in its constructor in order to specialize this class.

This class can also be treated as an interface by non-Bayesian Bandit solvers such as EpsilonGreedy. Such classes will need to over-ride all of the methods here.

Constructor Summary

Constructors

Constructor and Description
BayesianBandit()

Method Summary

Methods

Modifier and Type	Method and Description
boolean	addModelDistribution(AbstractBayesianDistribution distribution)
double	getMean(int k) Returns the mean of a particular distribution in the bandit
Iterator<AbstractBayesianDistribution>	iterator() Returns an iterator over a set of elements of type T.
List<Integer>	rank(int sampleSize) Samples probability estimates from each bandit and orders the bandits in increasing order.
int	sample() Samples probability estimates from each bandit and picks the apparent best
void	train(int bandit, double reward) Apply feedback to the bandit we chose.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

BayesianBandit

public BayesianBandit()

Method Detail

sample

public int sample()

Samples probability estimates from each bandit and picks the apparent best

Returns:

0 or 1 according to which bandit seems better

train

public void train(int bandit,
         double reward)

Apply feedback to the bandit we chose.

Parameters:

bandit - Which bandit got the impression

reward - Did it pay off?

addModelDistribution

public boolean addModelDistribution(AbstractBayesianDistribution distribution)

rank

public List<Integer> rank(int sampleSize)

Samples probability estimates from each bandit and orders the bandits in increasing order.

Parameters:

sampleSize - The number of bandits to sample.

Returns:

A list of the indexes of the bandits.

getMean

public double getMean(int k)

Returns the mean of a particular distribution in the bandit

iterator

public Iterator<AbstractBayesianDistribution> iterator()

Returns an iterator over a set of elements of type T.

Specified by:

iterator in interface Iterable<AbstractBayesianDistribution>

Returns:

an Iterator.