All Downloads are FREE. Search and download functionalities are using the official Maven repository.

org.cicirello.search.operators.permutations.BlockMoveMutation Maven / Gradle / Ivy

Go to download

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.permutations;

import org.cicirello.math.rand.RandomIndexer;
import org.cicirello.permutations.Permutation;
import org.cicirello.search.operators.IterableMutationOperator;
import org.cicirello.search.operators.MutationIterator;
import org.cicirello.search.operators.UndoableMutationOperator;

/**
 * This class implements a block move mutation on permutations, where one mutation consists in
 * removing a randomly chosen "block" (i.e., subsequence) and reinserting it at a different randomly
 * chosen index. The block move is chosen uniformly at random from among all possible block moves.
 *
 * 

A block move is sometimes also called a block transposition, and can be described equivalently * as swapping two adjacent blocks. Consider the permutation: p1 = [0, 1, 2, 3, 4, 5, 6, 7]. Now * consider a block move that consists in removing block [4, 5, 6] and reinserting it at index 2. * The result is p2 = [0, 1, 4, 5, 6, 2, 3, 7]. This can be described equivalently as swapping the * two adjacent blocks [4, 5, 6] and [2, 3]. This mutation operator is related to the {@link * BlockInterchangeMutation}, which swaps a pair of randomly selected non-overlapping blocks. * *

The runtime (worst case and average case) of both the {@link #mutate(Permutation) mutate} and * {@link #undo(Permutation) undo} methods is O(n), where n is the length of the permutation. The * worst case runtime occurs when the removed block is at one end of the permutation, and reinserted * at the opposite end, which causes all n permutation elements to move. On average, a block move * affects n/2 element locations. * * @author Vincent A. Cicirello, https://www.cicirello.org/ */ public class BlockMoveMutation implements UndoableMutationOperator, IterableMutationOperator { // needed to implement undo private final int[] indexes; /** Constructs a BlockMoveMutation mutation operator. */ public BlockMoveMutation() { indexes = new int[3]; } @Override public final void mutate(Permutation c) { if (c.length() >= 2) { generateIndexes(c.length(), indexes); c.removeAndInsert(indexes[1], indexes[2] - indexes[1] + 1, indexes[0]); } } @Override public final void undo(Permutation c) { c.removeAndInsert(indexes[0], indexes[2] - indexes[1] + 1, indexes[1]); } @Override public BlockMoveMutation split() { return new BlockMoveMutation(); } /** * {@inheritDoc} * *

The worst case runtime of the {@link MutationIterator#hasNext} and the {@link * MutationIterator#setSavepoint} methods of the {@link MutationIterator} created by this method * is O(1). The amortized runtime of the {@link MutationIterator#nextMutant} method is O(1). And * the worst case runtime of the {@link MutationIterator#rollback} method is O(n), where n is the * length of the Permutation. */ @Override public MutationIterator iterator(Permutation p) { return new BlockMoveIterator(p); } /* * This package access method allows the window limited version * implemented as a subclass to change how indexes are generated * without modifying the mutate method. */ void generateIndexes(int n, int[] indexes) { // Note 1: The nextIntTriple method returns 3 all different indexes, // but a removed block of length 1 would require 2 identical indexes. // To handle this, add 1 to n, and map an index beyond end of permutation // to the block length 1 case. // Note 2: Without loss of generality, the indexes are generated to // move the block earlier in the permutation. We can do this because // a "block move" essentially swaps two adjacent "blocks." RandomIndexer.nextIntTriple(n + 1, indexes, true); if (indexes[2] == n) indexes[2] = indexes[1]; } }





© 2015 - 2025 Weber Informatics LLC | Privacy Policy