org.cicirello.search.operators.UndoableMutationOperator 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

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.

There is a newer version: 7.0.1

Show newest version

/*
 * Chips-n-Salsa: A library of parallel self-adaptive local search algorithms.
 * Copyright (C) 2002-2021 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.operators;

/**
 * Implement the UndoableMutationOperator interface to implement a mutation operator for use in
 * simulated annealing, and other metaheuristics, that require a way to generate random neighbors of
 * a candidate solution, and which supports an undo method.
 *
 * The purpose of this subinterface is to enable efficient implementation of metaheuristics such
 * as simulated annealing, which iteratively generate random neighbors moving to some and not moving
 * to others. When the search chooses not to keep a neighbor, it needs an efficient way to return to
 * the previous state. Without an undo method, it would need to save a copy of the original.
 * Generating a copy of a candidate solution c is likely an operation whose cost is linear in the
 * size of c, while in many cases it may be possible to implement undo in constant time.
 *
 * 
If your mutation operator is one in which the inverse operation (i.e., the operation that
 * reverts an object to its previous state from prior to the mutation) can be implemented without
 * substantially affecting the runtime of the mutation itself, then implement the
 * UndoableMutationOperator interface.
 *
 * 
On the other hand, if implementing {@link #undo} would require significant added cost in
 * either time or memory to the {@link MutationOperator#mutate} method, then consider implementing
 * two versions of your mutation operator, one that implements MutationOperator and a second that
 * implements {@link UndoableMutationOperator}. In this way, you can use the first version with
 * metaheuristics that do not utilize the undo method, and the second for those that do.
 *
 * @param  The type of object used to represent candidate solutions to the problem.
 * @author Vincent A. Cicirello, https://www.cicirello.org/
 */
public interface UndoableMutationOperator extends MutationOperator {

  /**
   * Returns a candidate solution to its previous state prior to the most recent mutation performed.
   *
   * 
For example, consider the following. Let c' be the current state of c. Let c'' be the state
   * of c after mutate(c); If we then call undo(c), the state of c should revert back to c'.
   *
   * The behavior of undo is undefined if c is altered by some other process between the calls to
   * mutate and undo. The behavior is also undefined if a different candidate is given to undo then
   * the last given to mutate. For example, if the following two statements are executed, mutate(c);
   * undo(d);, the effect on d is undefined as it wasn't the most recently mutated candidate
   * solution.
   *
   * @param c The candidate solution to revert.
   */
  void undo(T c);

  @Override
  UndoableMutationOperator split();
}