Package etomica.integrator.mcmove

Class MCMoveStepTracker

java.lang.Object

etomica.integrator.mcmove.MCMoveTracker

etomica.integrator.mcmove.MCMoveStepTracker

public class MCMoveStepTracker
extends MCMoveTracker

This Class tracks acceptance statistics for an MCMove and also manages the move's step size. Given an acceptance target, it can adjust the step size (subject to min/max constraints) to achieve the desired acceptance ratio.

Field Summary

Fields

Modifier and Type	Field	Description
protected double	acceptanceTarget
protected long	adjustInterval
protected double	adjustStep
protected long	defaultAdjustInterval
protected double	defaultAdjustStep
protected long	lastAdjust
protected long	maxAdjustInterval
protected MCMoveStepDependent	mcMove
protected double	minAdjustStep
protected boolean	noisyAdjustment
protected boolean	noReset
protected boolean	tunable

Fields inherited from class etomica.integrator.mcmove.MCMoveTracker

chiSum, nAccept, nTrials

Constructor Summary

Constructors

Constructor	Description
MCMoveStepTracker()

Method Summary

Modifier and Type	Method	Description
protected void	adjustStepSize()
double	getAcceptanceTarget()
long	getAdjustInterval()
double	getAdjustStepSize()
boolean	getNoisyAdjustment()
boolean	getTunable()
void	resetAdjustStep()	Resets damped step adjustment.
void	setAcceptanceTarget(double target)	Sets the desired rate of acceptance (as a fraction between 0 and 1 inclusive) of trials of this move.
void	setAdjustInterval(long i)	Sets the interval between steps size adjustments.
void	setAdjustStepSize(double s)	Sets the step adjustment size.
void	setMaxAdjustInterval(long maxAdjustInterval)	The step size is adjusted every adjustInterval trials.
void	setMCMove(MCMoveStepDependent newMCMove)
void	setMinAdjustStep(double minAdjustStep)	When the step size is adjusted, it is adjusted by adjustStep.
void	setNoisyAdjustment(boolean isNoisy)
void	setTunable(boolean b)	Sets a flag to indicate whether tuning of the move is to be performed Tuning adjust the step size or other property of the move with the aim of achieving a trial acceptance rate that is near to some target value.
void	updateCounts(boolean moveWasAccepted, double chi)	Updates statistics regarding the acceptance rate of this move.

Methods inherited from class etomica.integrator.mcmove.MCMoveTracker

acceptanceProbability, acceptanceRatio, reset

Methods inherited from class java.lang.Object

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

Field Detail

mcMove

protected MCMoveStepDependent mcMove

acceptanceTarget

protected double acceptanceTarget

adjustInterval

protected long adjustInterval

maxAdjustInterval

protected long maxAdjustInterval

tunable

protected boolean tunable

lastAdjust

protected long lastAdjust

adjustStep

protected double adjustStep

minAdjustStep

protected double minAdjustStep

defaultAdjustStep

protected double defaultAdjustStep

defaultAdjustInterval

protected long defaultAdjustInterval

noisyAdjustment

protected boolean noisyAdjustment

noReset

protected boolean noReset

Constructor Detail

MCMoveStepTracker

public MCMoveStepTracker()

Method Detail

setMCMove

public void setMCMove(MCMoveStepDependent newMCMove)

updateCounts

public void updateCounts(boolean moveWasAccepted,
                         double chi)

Description copied from class: MCMoveTracker

Updates statistics regarding the acceptance rate of this move. This method should only be called by the Integrator after a move has been accepted or rejected.

Overrides:

updateCounts in class MCMoveTracker

Parameters:

moveWasAccepted - indicates whether the most recently attempted move was accepted

chi - indicates the probability with which the MC move was accepted.

adjustStepSize

protected void adjustStepSize()

resetAdjustStep

public void resetAdjustStep()

Resets damped step adjustment. The adjustment interval is reset to defaultAdjustInterval and the adjustment step is reset to defaultAdjustStep.

setMaxAdjustInterval

public void setMaxAdjustInterval(long maxAdjustInterval)

The step size is adjusted every adjustInterval trials. When the direction of the change oscillates, the adjustInterval is increased, but it will not be increased beyond maxAdjustInterval

Parameters:

maxAdjustInterval - the value beyond which adjustInterval will not be increased.

setMinAdjustStep

public void setMinAdjustStep(double minAdjustStep)

When the step size is adjusted, it is adjusted by adjustStep. When the direction of the change oscillates, the adjustStep is decreased, but it will not be decreased beyond minAdjustStep.

Parameters:

minAdjustStep - the value beyond which adjustStep will not be decreased.

getAcceptanceTarget

public final double getAcceptanceTarget()

Returns:

current value of the targeted rate of acceptance for this move.

setAcceptanceTarget

public final void setAcceptanceTarget(double target)

Sets the desired rate of acceptance (as a fraction between 0 and 1 inclusive) of trials of this move. IllegalArgumentException is thrown if given value is outside of acceptable range. The average acceptance ratio and probability and the step size adjustment parameters and convergence process are all reset.

getAdjustInterval

public final long getAdjustInterval()

setAdjustInterval

public void setAdjustInterval(long i)

Sets the interval between steps size adjustments. This also resets the acceptance averages and other step adjustment parameters to their default values.

getAdjustStepSize

public double getAdjustStepSize()

setAdjustStepSize

public void setAdjustStepSize(double s)

Sets the step adjustment size. The step size is multiplied by (1+s) when the acceptance ratio is too high and (1-s) when the step size is too low. This also resets the acceptance averages and other step adjustment parameters to their default values.

getTunable

public final boolean getTunable()

setTunable

public final void setTunable(boolean b)

Sets a flag to indicate whether tuning of the move is to be performed Tuning adjust the step size or other property of the move with the aim of achieving a trial acceptance rate that is near to some target value. Some moves (e.g., simple insertion trial) are inherently untunable, and will have false for this property.

getNoisyAdjustment

public boolean getNoisyAdjustment()

setNoisyAdjustment

public void setNoisyAdjustment(boolean isNoisy)