
org.cicirello.search.hc.AbstractHillClimber Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of chips-n-salsa Show documentation
Show all versions of chips-n-salsa Show documentation
Chips-n-Salsa is a Java library of customizable,
hybridizable, iterative, parallel, stochastic, and self-adaptive
local search algorithms. The library includes implementations of
several stochastic local search algorithms, including simulated
annealing, hill climbers, as well as constructive search algorithms
such as stochastic sampling. Chips-n-Salsa now also includes genetic
algorithms as well as evolutionary algorithms more generally. The
library very extensively supports simulated annealing. It includes
several classes for representing solutions to a variety of optimization
problems. For example, the library includes a BitVector class that
implements vectors of bits, as well as classes for representing
solutions to problems where we are searching for an optimal vector
of integers or reals. For each of the built-in representations, the
library provides the most common mutation operators for generating
random neighbors of candidate solutions, as well as common crossover
operators for use with evolutionary algorithms. Additionally, the
library provides extensive support for permutation optimization
problems, including implementations of many different mutation
operators for permutations, and utilizing the efficiently implemented
Permutation class of the JavaPermutationTools (JPT) library.
Chips-n-Salsa is customizable, making extensive use of Java's generic
types, enabling using the library to optimize other types of representations
beyond what is provided in the library. It is hybridizable, providing
support for integrating multiple forms of local search (e.g., using a hill
climber on a solution generated by simulated annealing), creating hybrid
mutation operators (e.g., local search using multiple mutation operators),
as well as support for running more than one type of search for the same
problem concurrently using multiple threads as a form of algorithm portfolio.
Chips-n-Salsa is iterative, with support for multistart metaheuristics,
including implementations of several restart schedules for varying the run
lengths across the restarts. It also supports parallel execution of multiple
instances of the same, or different, stochastic local search algorithms for
an instance of a problem to accelerate the search process. The library
supports self-adaptive search in a variety of ways, such as including
implementations of adaptive annealing schedules for simulated annealing,
such as the Modified Lam schedule, implementations of the simpler annealing
schedules but which self-tune the initial temperature and other parameters,
and restart schedules that adapt to run length.
/*
* Chips-n-Salsa: A library of parallel self-adaptive local search algorithms.
* Copyright (C) 2002-2022 Vincent A. Cicirello
*
* This file is part of Chips-n-Salsa (https://chips-n-salsa.cicirello.org/).
*
* Chips-n-Salsa is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* Chips-n-Salsa is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see .
*/
package org.cicirello.search.hc;
import org.cicirello.search.Metaheuristic;
import org.cicirello.search.ProgressTracker;
import org.cicirello.search.SimpleLocalMetaheuristic;
import org.cicirello.search.SolutionCostPair;
import org.cicirello.search.operators.Initializer;
import org.cicirello.util.Copyable;
/**
* This class serves as an abstract base class for the Hill Climbing implementations, including the
* common functionality.
*
* @param The type of object under optimization.
* @author Vincent A. Cicirello, https://www.cicirello.org/
*/
abstract class AbstractHillClimber>
implements Metaheuristic, SimpleLocalMetaheuristic {
private final Initializer initializer;
private ProgressTracker tracker;
private long neighborCount;
/**
* Constructs a hill climber object.
*
* @param initializer The source of random initial states for each hill climb.
* @param tracker A ProgressTracker object, which is used to keep track of the best solution found
* during the run, the time when it was found, and other related data.
* @throws NullPointerException if any of the parameters are null.
*/
AbstractHillClimber(Initializer initializer, ProgressTracker tracker) {
if (initializer == null || tracker == null) {
throw new NullPointerException();
}
this.initializer = initializer;
this.tracker = tracker;
}
/*
* package-private copy constructor in support of the split method.
* note: copies references to thread-safe components, and splits potentially non-threadsafe components
*/
AbstractHillClimber(AbstractHillClimber other) {
// this one must be shared.
tracker = other.tracker;
// split: not threadsafe
initializer = other.initializer.split();
// use default of 0 for this one: neighborCount
}
@Override
public final SolutionCostPair optimize() {
if (tracker.didFindBest() || tracker.isStopped()) return null;
neighborCount++;
return climbOnce(initializer.createCandidateSolution());
}
@Override
public final SolutionCostPair optimize(T start) {
if (tracker.didFindBest() || tracker.isStopped()) return null;
return climbOnce(start.copy());
}
/**
* Executes multiple restarts of the hill climber. Each restart begins from a new random starting
* solution. Returns the best solution across the restarts.
*
* @param numRestarts The number of restarts of the hill climber.
* @return The best solution of this set of restarts, which may or may not be the same as the
* solution contained in this hill climber's {@link org.cicirello.search.ProgressTracker
* ProgressTracker}, which contains the best of all runs across all calls to the various
* optimize methods. Returns null if no runs executed, such as if the ProgressTracker already
* contains the theoretical best solution.
*/
@Override
public final SolutionCostPair optimize(int numRestarts) {
if (tracker.didFindBest() || tracker.isStopped()) return null;
SolutionCostPair best = null;
for (int i = 0; i < numRestarts && !tracker.didFindBest() && !tracker.isStopped(); i++) {
SolutionCostPair current = climbOnce(initializer.createCandidateSolution());
neighborCount++;
if (best == null || current.compareTo(best) < 0) best = current;
}
return best;
}
@Override
public final ProgressTracker getProgressTracker() {
return tracker;
}
@Override
public final void setProgressTracker(ProgressTracker tracker) {
if (tracker != null) this.tracker = tracker;
}
/**
* Gets the total run length, where run length is number of candidate solutions generated by the
* hill climber. This is the total run length across all calls to the search.
*
* @return the total number of candidate solutions generated by the search, across all calls to
* the various optimize methods.
*/
@Override
public final long getTotalRunLength() {
return neighborCount;
}
@Override
public abstract AbstractHillClimber split();
final SolutionCostPair reportSingleClimbStatus(
int currentCost, T current, boolean isMinCost, long neighborCountIncrement) {
neighborCount = neighborCount + neighborCountIncrement;
// update tracker
if (currentCost < tracker.getCost()) {
tracker.update(currentCost, current, isMinCost);
}
return new SolutionCostPair(current, currentCost, isMinCost);
}
final SolutionCostPair reportSingleClimbStatus(
double currentCost, T current, boolean isMinCost, long neighborCountIncrement) {
neighborCount = neighborCount + neighborCountIncrement;
// update tracker
if (currentCost < tracker.getCostDouble()) {
tracker.update(currentCost, current, isMinCost);
}
return new SolutionCostPair(current, currentCost, isMinCost);
}
interface OneClimb> {
SolutionCostPair climb(T current);
}
abstract SolutionCostPair climbOnce(T current);
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy