org.terrier.matching.JforestsModelMatching Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of terrier-learning Show documentation
The newest version!
/*
 * Terrier - Terabyte Retriever 
 * Webpage: http://terrier.org 
 * Contact: terrier{a.}dcs.gla.ac.uk
 * University of Glasgow - School of Computing Science
 * http://www.gla.ac.uk/
 * 
 * The contents of this file are subject to the Mozilla Public License
 * Version 1.1 (the "License"); you may not use this file except in
 * compliance with the License. You may obtain a copy of the License at
 * http://www.mozilla.org/MPL/
 *
 * Software distributed under the License is distributed on an "AS IS"
 * basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
 * the License for the specific language governing rights and limitations
 * under the License.
 *
 * The Original Code is JforestsModelMatching.java.
 *
 * The Original Code is Copyright (C) 2004-2020 the University of Glasgow.
 * All Rights Reserved.
 *
 * Contributor(s):
 *   Craig Macdonald 
 */

package org.terrier.matching;

import java.io.File;
import java.util.Arrays;

import org.terrier.learning.FeaturedResultSet;
import org.terrier.structures.Index;
import org.terrier.utility.ApplicationSetup;

import edu.uci.jforests.dataset.BitNumericArray;
import edu.uci.jforests.dataset.ByteNumericArray;
import edu.uci.jforests.dataset.Feature;
import edu.uci.jforests.dataset.NullNumericArray;
import edu.uci.jforests.dataset.NumericArray;
import edu.uci.jforests.dataset.RankingDataset;
import edu.uci.jforests.dataset.ShortNumericArray;
import edu.uci.jforests.input.FeatureAnalyzer;
import edu.uci.jforests.learning.LearningUtils;
import edu.uci.jforests.learning.trees.Ensemble;
import edu.uci.jforests.learning.trees.Tree;
import edu.uci.jforests.learning.trees.decision.DecisionTree;
import edu.uci.jforests.learning.trees.regression.RegressionTree;
import edu.uci.jforests.sample.RankingSample;
import edu.uci.jforests.sample.Sample;
import gnu.trove.TIntIntHashMap;

/** Applies a Jforests regression tree learned model to a {@link FeaturedResultSet}. Learned model files are generated by Jforests.
 * If you use this class, you are implicitly using the jforests library, and the Jforests citation policy applies.
 * Properties
 * 
 * fat.matching.learned.jforest.model - filename of the ensemble model generated by jforests
 * fat.matching.learned.jforest.statistics - filename of the feature statistics file generated by jforests
 * 
 * @author Craig Macdonald
 * @since 4.0
 */
public class JforestsModelMatching extends LearnedModelMatching {

	final Ensemble ensemble = new Ensemble();
	final FeatureAnalyzer featureAnalyzer = new FeatureAnalyzer();
	
	public JforestsModelMatching(Index _index, Matching _parent) throws Exception {
		super(_index, _parent);
		loadModel(ApplicationSetup.getProperty("fat.matching.learned.jforest.model", null));
	}

	public JforestsModelMatching(Matching _parent) throws Exception {
		super(_parent);
		loadModel(ApplicationSetup.getProperty("fat.matching.learned.jforest.model", null));
	}
	
	public JforestsModelMatching(Matching _parent, String modelFilename) throws Exception {
		super(_parent);
		loadModel(modelFilename);
	}
	
	public JforestsModelMatching(Index _index, Matching _parent, String modelFilename) throws Exception {
		super(_parent);
		loadModel(modelFilename);
	}
	
	public JforestsModelMatching(Index _index, Matching _parent, String modelFilename, Class treeClass) throws Exception {
		super(_parent);
		loadModel(modelFilename, treeClass);
	}
	
	protected void loadModel(String model_filename) throws Exception
	{
		final boolean regression = true;
		loadModel(model_filename, regression ? RegressionTree.class : DecisionTree.class);
	}
	
	protected void loadModel(String model_filename, Class treeClass) throws Exception
	{
		if (model_filename == null)
			throw new IllegalArgumentException("model_filename not specified, perhaps you did not specify property fat.matching.learned.jforest.model ?" );
		ensemble.loadFromFile(treeClass, new File(model_filename));
		final String featureStats_filename = 
				ApplicationSetup.getProperty("fat.matching.learned.jforest.statistics", model_filename+".features");
		featureAnalyzer.loadFeaturesFromFile(featureStats_filename);
	}
	
	protected RankingDataset makeDataset(int N, int featureCount, double[][] doubleFeatures)
	{
		//doubleFeatures: indexed by feature then document
		//intFeatures: indexed by document then feature
		final int[][] intFeatures = new int[N][featureCount];
        TIntIntHashMap[] _valueHashMap = new TIntIntHashMap[featureCount];
        for (int f = 0; f < featureCount; f++) {
                _valueHashMap[f] = new TIntIntHashMap();
        }
        
        //we are also rotating here.
        for(int d=0;d