edu.oregonstate.eecs.mcplan.search.fsss.EpsilonGreedyRefinementOrder.java Source code

Introduction

Here is the source code for edu.oregonstate.eecs.mcplan.search.fsss.EpsilonGreedyRefinementOrder.java

Source

/* LICENSE
Copyright (c) 2013-2016, Jesse Hostetler (jessehostetler@gmail.com)
All rights reserved.
    
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:
    
1. Redistributions of source code must retain the above copyright notice,
   this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice,
   this list of conditions and the following disclaimer in the documentation
   and/or other materials provided with the distribution.
    
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" 
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/

/**
 * 
 */
package edu.oregonstate.eecs.mcplan.search.fsss;

import java.util.ArrayList;
import java.util.Iterator;

import org.apache.commons.math3.random.RandomGenerator;

import edu.oregonstate.eecs.mcplan.State;
import edu.oregonstate.eecs.mcplan.VirtualConstructor;

/**
 * @author jhostetler
 *
 */
public class EpsilonGreedyRefinementOrder<S extends State, A extends VirtualConstructor<A>>
        extends RefinementOrderBase<S, A> {
    public static class Factory<S extends State, A extends VirtualConstructor<A>>
            implements RefinementOrderBase.Factory<S, A> {
        private final RandomGenerator rng;
        private final double epsilon;
        private final SubtreeRefinementOrder.Factory<S, A> subtree_factory;

        public Factory(final RandomGenerator rng, final double epsilon,
                final SubtreeRefinementOrder.Factory<S, A> subtree_factory) {
            this.rng = rng;
            this.epsilon = epsilon;
            this.subtree_factory = subtree_factory;
        }

        @Override
        public String toString() {
            return "EpsilonGreedy(" + epsilon + "; " + subtree_factory + ")";
        }

        @Override
        public RefinementOrderBase<S, A> create(final FsssParameters parameters, final FsssModel<S, A> model,
                final FsssAbstractStateNode<S, A> root) {
            final ArrayList<SubtreeRefinementOrder<S, A>> subtrees = new ArrayList<SubtreeRefinementOrder<S, A>>();
            for (final FsssAbstractActionNode<S, A> aan : root.successors()) {
                subtrees.add(subtree_factory.create(parameters, model, aan));
            }
            assert (subtrees.size() == root.nsuccessors());
            return new EpsilonGreedyRefinementOrder<S, A>(rng, epsilon, root, subtrees);
        }
    }

    // -----------------------------------------------------------------------

    private final RandomGenerator rng;
    private final double epsilon;

    public EpsilonGreedyRefinementOrder(final RandomGenerator rng, final double epsilon,
            final FsssAbstractStateNode<S, A> root, final ArrayList<SubtreeRefinementOrder<S, A>> subtrees) {
        super(root, subtrees);
        this.rng = rng;
        this.epsilon = epsilon;
    }

    @Override
    public boolean isClosed() {
        return closed;
    }

    @Override
    protected SubtreeRefinementOrder<S, A> chooseSubtree() {
        if (rng.nextDouble() > epsilon) {
            // Greedy choice of the *2nd-best* action
            final FsssAbstractActionNode<S, A> aan = root.astar_random();
            assert (aan != null);

            final ArrayList<SubtreeRefinementOrder<S, A>> astar = new ArrayList<SubtreeRefinementOrder<S, A>>();
            double ustar = -Double.MAX_VALUE;
            for (final SubtreeRefinementOrder<S, A> subtree : subtrees.values()) {
                final FsssAbstractActionNode<S, A> ai = subtree.rootAction();
                if (ai == aan) {
                    // Skip the optimal subtree
                    continue;
                }
                final double u = ai.U();
                if (u > ustar) {
                    ustar = u;
                    astar.clear();
                    astar.add(subtree);
                } else if (u >= ustar) {
                    astar.add(subtree);
                }
            }

            final SubtreeRefinementOrder<S, A> subtree = astar.get(rng.nextInt(astar.size()));
            //         System.out.println( "\tEpsilonGreedyRefinementOrder: greedy choice " + subtree.rootAction() );
            return subtree;
        } else {
            // Random choice
            final int choice = rng.nextInt(subtrees.size());
            final Iterator<SubtreeRefinementOrder<S, A>> itr = subtrees.values().iterator();
            for (int i = 0; i < choice; ++i) {
                itr.next();
            }
            final SubtreeRefinementOrder<S, A> subtree = itr.next();
            //         System.out.println( "\tEpsilonGreedyRefinementOrder: random subtree " + subtree.rootAction() );
            return subtree;
        }
    }
}