de.uniba.sme.bambirds.level_selection.Action Class Reference

Collaboration diagram for de.uniba.sme.bambirds.level_selection.Action:

Classes
enum	Strategy

Public Member Functions
	Action (double epsilon, Strategy strategy, long timeLimit, int numberOfLevels)
	Action (double epsilon, Strategy strategy, long timeLimit, int numberOfLevels, long seed)
int	nextLevel (Map< Integer, Double > probabilities, long timeLeft)
long	getTimeLimit ()
void	setTimeLimit (long timeLimit)

Private Member Functions
int	oldSelection ()
int	greedySelection (Map< Integer, Double > probabilities)
int	epsilonGreedySelection (Map< Integer, Double > probabilities)
int	epsilonAdaptiveSelection (Map< Integer, Double > probabilities, long timeLeft)
int	randomSelection (Map< Integer, Double > probabilities)

Private Attributes
double	epsilonInit
Strategy	strategy
Random	random
long	timeLimit
int	numberOfLevels
int	currentLevel
int	numberOfTimesPlayedTheMost = -1
int	numberOfTimesPlayedMinimum = 1000
boolean	allLostLevelsPlayedTwice = false
int	lastLostLevel = -1

Static Private Attributes
static final Logger	log = LogManager.getLogger(Action.class)

Detailed Description

Class for selecting a Level using a specific Strategy and the probability distribution calculated by Decision

Constructor & Destructor Documentation

Action() [1/2]

de.uniba.sme.bambirds.level_selection.Action.Action	(	double	epsilon,
		Strategy	strategy,
		long	timeLimit,
		int	numberOfLevels
	)

Create a new Action Object

Parameters

epsilon	initial value for epsilon algorithms
strategy	Type of strategy used to select next level
timeLimit	Maximum time in the Competition
numberOfLevels	The number of Levels the agent is playing

Action() [2/2]

de.uniba.sme.bambirds.level_selection.Action.Action	(	double	epsilon,
		Strategy	strategy,
		long	timeLimit,
		int	numberOfLevels,
		long	seed
	)

Create a new Action Object

Parameters

epsilon	initial value for epsilon algorithms
strategy	Type of strategy used to select next level
timeLimit	Maximum time in the Competition
numberOfLevels	The number of Levels the agent is playing
seed	The seed for the random number generator

Member Function Documentation

epsilonAdaptiveSelection()

int de.uniba.sme.bambirds.level_selection.Action.epsilonAdaptiveSelection ( Map< Integer, Double >  probabilities, long  timeLeft  )

private

Select the level with the highest probability depending on how much time is left and how good the score and win predictions are.
Otherwise randomSelection

Parameters

probabilities	for each possible level
timeLeft	remaining time

Returns

the next level to play

epsilonGreedySelection()

int de.uniba.sme.bambirds.level_selection.Action.epsilonGreedySelection ( Map< Integer, Double >  probabilities )

private

Select the level with the highest probability with probability 1 - epsilon.

Otherwise randomSelection

Parameters

probabilities

for each possible level

Returns

the next level to play

greedySelection()

int de.uniba.sme.bambirds.level_selection.Action.greedySelection ( Map< Integer, Double >  probabilities )

private

Select the level with the highest probability

Parameters

probabilities

for each possible level

Returns

the next level to play

nextLevel()

int de.uniba.sme.bambirds.level_selection.Action.nextLevel	(	Map< Integer, Double >	probabilities,
		long	timeLeft
	)

Calculate the next Level to be played depending on the strategy.

Parameters

probabilities	for each possible level
timeLeft	remaining time

Returns

the next level to play

oldSelection()

int de.uniba.sme.bambirds.level_selection.Action.oldSelection ( )

private

The old Level Selection, where first all not won levels will be attempted 3 times. Then if less than 15 percent have been lost, the Level with the highest possible improvement based on the maximum possible points will be selected

Returns

the next level to play

randomSelection()

int de.uniba.sme.bambirds.level_selection.Action.randomSelection ( Map< Integer, Double >  probabilities )

private

Randomly select a level depending on its probability

Parameters

probabilities

for each possible level

Returns

the next level to play

---

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

• Action.java