org.cicirello.search.operators.bits.BitFlipMutation 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.bits;

import org.cicirello.search.operators.UndoableMutationOperator;
import org.cicirello.search.representations.BitVector;

/**
 * This class implements Bit Flip Mutation, the mutation operator commonly used in genetic
 * algorithms, but which can also be used with other metaheuristic search algorithms such as
 * simulated annealing to generate random neighbors.
 *
 * In a bit flip mutation, each bit of the BitVector is flipped with probability M, known as the
 * mutation rate. Flipping a bit means changing it to 0 if it is currently a 1, or changing it to 1
 * if it is currently a 0. If the length of the BitVector is N, then the expected number of bits
 * flipped during a single call to the {@link #mutate} method is NM. However, there is no guarantee
 * that any bits will be flipped by a call to the {@link #mutate} method. This behavior is fine for
 * genetic algorithms, but may be less than desirable for other metaheuristics, such as those that
 * operate on a single candidate solution rather than a population of them. If you have need for a
 * mutation operator for BitVectors that guarantees that all calls to the {@link #mutate} method
 * will change the BitVector, then consider using the {@link DefiniteBitFlipMutation} class instead.
 *
 * @author Vincent A. Cicirello, https://www.cicirello.org/
 */
public final class BitFlipMutation implements UndoableMutationOperator {

  private final double m;
  private BitVector bitMask;

  /**
   * Constructs a BitFlipMutation operator with a specified mutation rate.
   *
   * @param m The mutation rate, which is the probability of flipping any individual bit. The
   *     expected number of bits flipped during a call to the {@link #mutate} method is m*N where N
   *     is the length of the mutated BitVector. There is no guarantee that any bits will be flipped
   *     during a mutation (e.g., if m is close to 0).
   * @throws IllegalArgumentException if m ≤ 0 or if m ≥ 1.
   */
  public BitFlipMutation(double m) {
    if (m <= 0 || m >= 1) throw new IllegalArgumentException("m constrained by: 0.0 < m < 1.0");
    this.m = m;
  }

  /*
   * internal copy constructor
   */
  private BitFlipMutation(BitFlipMutation other) {
    m = other.m;
    // deliberately don't copy bitMask (each instance needs to maintain its own for undo)
  }

  @Override
  public void mutate(BitVector c) {
    bitMask = new BitVector(c.length(), m);
    c.xor(bitMask);
  }

  @Override
  public void undo(BitVector c) {
    if (bitMask != null) {
      c.xor(bitMask);
    }
  }

  @Override
  public BitFlipMutation split() {
    return new BitFlipMutation(this);
  }
}