Package edu.cnu.mdi.sim.simanneal.heuristics

Class EnergyDistributionHeuristic<S extends AnnealingSolution<S>>

java.lang.Object

edu.cnu.mdi.sim.simanneal.heuristics.EnergyDistributionHeuristic<S>

Type Parameters:

S - the concrete solution type for the annealing problem

All Implemented Interfaces:

TemperatureHeuristic<S>

public class EnergyDistributionHeuristic<S extends AnnealingSolution<S>>
extends Object
implements TemperatureHeuristic<S>

Estimates an initial temperature for simulated annealing by sampling the distribution of energies from random solutions.

This heuristic draws a number of independent random solutions from the AnnealingProblem, computes their energies, and uses robust statistics to estimate a characteristic energy scale ΔE. The initial temperature T0 is then chosen so that a typical uphill move of size ΔE would be accepted with a specified probability p:

     p ≈ exp(-ΔE / T0)  ⇒  T0 = -ΔE / ln(p)
 

Robust statistics

Rather than relying on the mean and standard deviation (which can be highly sensitive to outliers), this heuristic uses:

• Median of the sampled energies as a location estimate
• Median Absolute Deviation (MAD) as a robust scale estimate

The MAD is converted to a "sigma-like" scale using the standard normal consistency factor 1.4826. If the energy landscape is effectively flat (MAD ≈ 0), the heuristic falls back to an interquartile-range-based estimate.

Intended use

This heuristic is designed to be used during simulation initialization to automatically choose a reasonable starting temperature without manual tuning. It is problem-agnostic and works well for a wide range of combinatorial and continuous optimization problems.

Constructor Summary

Constructors

Constructor	Description
EnergyDistributionHeuristic(int samples, double targetAcceptance, double minT0)	Construct a new energy-distribution-based temperature heuristic.

Method Summary

Modifier and Type	Method	Description
InitialTemperature	estimate(AnnealingProblem<S> problem, Random rng)	Estimate an initial temperature by sampling the energy distribution of random solutions.

Methods inherited from class java.lang.Object

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

Constructor Details

EnergyDistributionHeuristic

public EnergyDistributionHeuristic(int samples,
 double targetAcceptance,
 double minT0)

Construct a new energy-distribution-based temperature heuristic.

Parameters:

samples - number of random solutions to sample (must be ≥ 10)

targetAcceptance - desired acceptance probability for a typical uphill move (must lie strictly in (0,1))

minT0 - minimum allowed initial temperature (non-negative)

Throws:

IllegalArgumentException - if samples < 10 or targetAcceptance ∉ (0,1)

Method Details

estimate

public InitialTemperature estimate(AnnealingProblem<S> problem,
 Random rng)

Estimate an initial temperature by sampling the energy distribution of random solutions.

Specified by:

estimate in interface TemperatureHeuristic<S extends AnnealingSolution<S>>

Parameters:

problem - the annealing problem providing random solutions and an energy function

rng - source of randomness

Returns:

an InitialTemperature record containing the estimated temperature and diagnostic statistics (median, MAD, sample count)