Homepage Demos Overview Downloads Tutorials Reference
Credits
Main Page | Namespace List | Class Hierarchy | Alphabetical List | Compound List | File List | Namespace Members | Compound Members | File Members | Related Pages | Search

karmedbanditExp3 Class Reference

#include <karmedbandit.h>

List of all members.

Detailed Description

Makes decisions regarding an adversarial k-armed bandit.

Uses algorithms described in: The non-stochastic multi-armed bandit problem Auer, Cesa-Bianchi, Freund, and Schapire October 14, 2002

Definition at line 15 of file karmedbandit.h.

Public Member Functions

 karmedbanditExp3 (unsigned int k, double gammap)
 constructor, pass the number of arms

unsigned int decide ()
 returns the next choice, [0:k-1]

void reward (bool r)
 call this if you want to reward (r==true) or penalize (r==false) the previous decision

void reset ()
 resets weights

double getGamma ()
 gets gamma parameter

void setGamma (double gammap)
 sets gamma parameter

unsigned int getK ()
 gets k parameter


Protected Attributes

std::vector< double > w
 the weights

double lastp
 prob of last choice

unsigned int last
 the last choice

double g
 gamma

Constructor & Destructor Documentation

karmedbanditExp3::karmedbanditExp3 (  unsigned int  k,
double  gammap
)  [inline]
 

constructor, pass the number of arms

Definition at line 18 of file karmedbandit.h.

References g, last, lastp, and w.

Member Function Documentation

unsigned int karmedbanditExp3::decide (   )  [inline]
 

returns the next choice, [0:k-1]

Definition at line 23 of file karmedbandit.h.

References g, last, lastp, and w.

double karmedbanditExp3::getGamma (   )  [inline]
 

gets gamma parameter

Definition at line 65 of file karmedbandit.h.

References g.

unsigned int karmedbanditExp3::getK (   )  [inline]
 

gets k parameter

Definition at line 69 of file karmedbandit.h.

References w.

void karmedbanditExp3::reset (   )  [inline]
 

resets weights

Definition at line 60 of file karmedbandit.h.

References w.

void karmedbanditExp3::reward (  bool  r  )  [inline]
 

call this if you want to reward (r==true) or penalize (r==false) the previous decision

Definition at line 52 of file karmedbandit.h.

References g, last, lastp, and w.

void karmedbanditExp3::setGamma (  double  gammap  )  [inline]
 

sets gamma parameter

Definition at line 67 of file karmedbandit.h.

References g.

Member Data Documentation

double karmedbanditExp3::g [protected]
 

gamma

Definition at line 74 of file karmedbandit.h.

unsigned int karmedbanditExp3::last [protected]
 

the last choice

Definition at line 73 of file karmedbandit.h.

double karmedbanditExp3::lastp [protected]
 

prob of last choice

Definition at line 72 of file karmedbandit.h.

std::vector<double> karmedbanditExp3::w [protected]
 

the weights

Definition at line 71 of file karmedbandit.h.

The documentation for this class was generated from the following file:
Tekkotsu v1.4
 		Generated Sat Jul 19 00:09:00 2003 by Doxygen 1.3.2