jp.ac.kobe_u.cs.cream.TabooSearch Maven / Gradle / Ivy

Go to download
/*
 * @(#)TabooSearch.java
 */
package jp.ac.kobe_u.cs.cream;

import java.util.Iterator;

/**
 * Taboo Search. Only useful for the optimization problems containing {@link Serialized} constraints
 * in the current implementation. The default exchangeRate value for {@link ParallelSolver} is 0.8.
 *
 * @since 1.0
 * @version 1.4
 * @author Naoyuki Tamura ([email protected])
 */
public class TabooSearch extends LocalSearch {
  /** The length of the taboo list. The default value is 16. */
  public int tabooLength = 16;

  private Operation[] taboo = null;

  private int taboo_i;

  public TabooSearch(Network network) {
    this(network, DEFAULT, null);
  }

  public TabooSearch(Network network, int option) {
    this(network, option, null);
  }

  public TabooSearch(Network network, String name) {
    this(network, DEFAULT, name);
  }

  public TabooSearch(Network network, int option, String name) {
    super(network, option, name);
    setExchangeRate(0.8);
  }

  /** Allocates a new taboo list. */
  public void clearTaboo() {
    taboo = new Operation[tabooLength];
    for (int i = 0; i < taboo.length; i++) {
      taboo[i] = null;
    }
    taboo_i = 0;
  }

  /**
   * Returns true when the operation op is a taboo.
   *
   * @param op the operation to be checked
   * @return true when op is a taboo
   */
  public boolean isTaboo(Operation op) {
    if (taboo == null) return false;
    for (int i = 0; i < taboo.length; i++) {
      if (taboo[i] != null && op.isTaboo(taboo[i])) return true;
    }
    return false;
  }

  /**
   * Adds the operation op to the taboo list.
   *
   * @param op the operation to be added
   */
  public void addTaboo(Operation op) {
    taboo[taboo_i] = op;
    taboo_i = (taboo_i + 1) % taboo.length;
  }

  @Override
  public void startSearch() {
    super.startSearch();
    clearTaboo();
  }

  @Override
  public void nextSearch() {
    Operation locallyBestOp = null;
    Solution locallyBestSol = null;
    int locallyBest = IntDomain.MAX_VALUE;
    solution = getCandidate();
    if (solution == null) return;
    Code code = solution.getCode();
    while (!isAborted()) {
      Iterator ops = code.operations().iterator();
      while (ops.hasNext() && !isAborted()) {
        Operation op = ops.next();
        if (isTaboo(op)) continue;
        code.setTo(network);
        op.applyTo(network);
        Solution sol = solver.findBest(iterationTimeout);
        if (sol == null) {
          continue;
        }
        int value = sol.getObjectiveIntValue();
        if (!isBetter(value, locallyBest)) {
          continue;
        }
        locallyBest = value;
        locallyBestOp = op;
        locallyBestSol = sol;
      }
      if (locallyBestOp != null) {
        break;
      }
      clearTaboo();
    }
    if (isAborted()) {
      return;
    }
    if (locallyBestOp != null) {
      code.setTo(network);
      locallyBestOp.applyTo(network);
      solution = locallyBestSol;
      addTaboo(locallyBestOp);
    }
  }
}