jcgp.backend.statistics

Class StatisticsLogger

java.lang.Object

jcgp.backend.statistics.StatisticsLogger

public class StatisticsLogger
extends java.lang.Object

This is a utility class for logging experiment statistics when doing multiple runs.

Information about each run is added via the logRun() method. The many getters can be used to obtain statistics about the logged runs, such as success rate and average fitness.

JCGP uses this class to perform its logging and print out experiment data at the end.

Author:

Eduardo Pedroni

Constructor Summary

Constructors

Constructor and Description
StatisticsLogger() Create a new statistics logger, use this when resetting is necessary.

Method Summary

Modifier and Type	Method and Description
double	getAverageActiveNodes() Averages the number of active nodes in the best chromosomes obtained across all runs.
double	getAverageActiveNodesStdDev() Calculates the standard deviation of the number of active nodes in the best solution in each run.
double	getAverageFitness() Averages the best fitness obtained in each run.
double	getAverageFitnessStdDev() Calculates the standard deviation of the best fitness obtained in each run.
double	getAverageGenerations() Calculates the average generation out of all runs.
double	getAverageGenerationsStdDev() Calculates the standard deviation of the average number of generations in each run.
double	getAverageSuccessfulGenerations() Calculates the average generation out of successful runs only.
double	getAverageSuccessfulGenerationsStdDev() Calculates the standard deviation of the average number of generations in each run where a perfect solution was found.
double	getHighestFitness()
double	getLowestFitness()
int	getSuccessfulRuns()
double	getSuccessRate() Calculates the ratio of successful runs (runs where a perfect solution was found) to total number of runs.
void	logRun(int generation, double fitness, int active, boolean successful) Log a new run.

Methods inherited from class java.lang.Object

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

Constructor Detail

StatisticsLogger

public StatisticsLogger()

Create a new statistics logger, use this when resetting is necessary.

Method Detail

logRun

public void logRun(int generation,
                   double fitness,
                   int active,
                   boolean successful)

Log a new run. Calling any of the statistics getters will now take this logged run into account as well as all previously logged runs.

Parameters:

generation - the last generation when improvement occurred.

fitness - the best fitness achieved in the run.

active - the number of active nodes in the best chromosome found.

successful - true if a perfect solution was found, false if otherwise.

getAverageFitness

public double getAverageFitness()

Averages the best fitness obtained in each run.

Returns:

the average fitness.

getAverageFitnessStdDev

public double getAverageFitnessStdDev()

Calculates the standard deviation of the best fitness obtained in each run.

Returns:

the standard deviation of average fitnesses.

getAverageActiveNodes

public double getAverageActiveNodes()

Averages the number of active nodes in the best chromosomes obtained across all runs.

Returns:

the average number of active nodes.

getAverageActiveNodesStdDev

public double getAverageActiveNodesStdDev()

Calculates the standard deviation of the number of active nodes in the best solution in each run.

Returns:

the standard deviation of active node counts.

getAverageGenerations

public double getAverageGenerations()

Calculates the average generation out of all runs. The generation value in each run corresponds to the last generation in which improvement happened.

Note that this method includes runs where no perfect solution was found. For the average number of generations for perfect solutions only, use getAverageSuccessfulGenerations.

Returns:

the average number of generations.

getAverageGenerationsStdDev

public double getAverageGenerationsStdDev()

Calculates the standard deviation of the average number of generations in each run.

Returns:

the standard deviation of the number of generations.

getHighestFitness

public double getHighestFitness()

Returns:

the highest fitness across all runs.

getLowestFitness

public double getLowestFitness()

Returns:

the lowest fitness across all runs.

getSuccessfulRuns

public int getSuccessfulRuns()

Returns:

the number of runs in which a perfect solution was found.

getSuccessRate

public double getSuccessRate()

Calculates the ratio of successful runs (runs where a perfect solution was found) to total number of runs. A double-precision value between 0 and 1 is returned, where 0 means 0% success rate and 1 means 100% success rate.

Returns:

the success rate across all runs.

getAverageSuccessfulGenerations

public double getAverageSuccessfulGenerations()

Calculates the average generation out of successful runs only. The generation value in each successful run corresponds to the generation in which the perfect solution was found.

Returns:

the average number of generations for perfect solutions.

getAverageSuccessfulGenerationsStdDev

public double getAverageSuccessfulGenerationsStdDev()

Calculates the standard deviation of the average number of generations in each run where a perfect solution was found.

Returns:

the standard deviation of the number of generations in successful runs.