/**
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (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.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.mahout.cf.taste.impl.recommender.svd;
import java.util.Collections;
import java.util.List;
import java.util.Random;
import com.google.common.collect.Lists;
import org.apache.mahout.cf.taste.common.TasteException;
import org.apache.mahout.cf.taste.impl.common.FullRunningAverage;
import org.apache.mahout.cf.taste.impl.common.LongPrimitiveIterator;
import org.apache.mahout.cf.taste.impl.common.RunningAverage;
import org.apache.mahout.cf.taste.model.DataModel;
import org.apache.mahout.cf.taste.model.Preference;
import org.apache.mahout.common.RandomUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
/** Calculates the SVD using an Expectation Maximization algorithm. */
public final class ExpectationMaximizationSVDFactorizer extends AbstractFactorizer {
private static final Logger log = LoggerFactory.getLogger(ExpectationMaximizationSVDFactorizer.class);
private final double learningRate;
/** Parameter used to prevent overfitting. 0.02 is a good value. */
private final double preventOverfitting;
/** number of features used to compute this factorization */
private final int numFeatures;
/** number of iterations */
private final int numIterations;
private final double randomNoise;
/** user singular vectors */
private double[][] leftVectors;
/** item singular vectors */
private double[][] rightVectors;
private final DataModel dataModel;
private List<SVDPreference> cachedPreferences;
private double defaultValue;
private double interval;
public ExpectationMaximizationSVDFactorizer(DataModel dataModel,
int numFeatures,
int numIterations) throws TasteException {
// use the default parameters from the old SVDRecommender implementation
this(dataModel, numFeatures, 0.005, 0.02, 0.005, numIterations);
}
public ExpectationMaximizationSVDFactorizer(DataModel dataModel,
int numFeatures,
double learningRate,
double preventOverfitting,
double randomNoise,
int numIterations) throws TasteException {
super(dataModel);
this.dataModel = dataModel;
this.numFeatures = numFeatures;
this.numIterations = numIterations;
this.learningRate = learningRate;
this.preventOverfitting = preventOverfitting;
this.randomNoise = randomNoise;
}
@Override
public Factorization factorize() throws TasteException {
Random random = RandomUtils.getRandom();
leftVectors = new double[dataModel.getNumUsers()][numFeatures];
rightVectors = new double[dataModel.getNumItems()][numFeatures];
double average = getAveragePreference();
double prefInterval = dataModel.getMaxPreference() - dataModel.getMinPreference();
defaultValue = Math.sqrt((average - prefInterval * 0.1) / numFeatures);
interval = prefInterval * 0.1 / numFeatures;
for (int feature = 0; feature < numFeatures; feature++) {
for (int userIndex = 0; userIndex < dataModel.getNumUsers(); userIndex++) {
leftVectors[userIndex][feature] = defaultValue + (random.nextDouble() - 0.5) * interval * randomNoise;
}
for (int itemIndex = 0; itemIndex < dataModel.getNumItems(); itemIndex++) {
rightVectors[itemIndex][feature] = defaultValue + (random.nextDouble() - 0.5) * interval * randomNoise;
}
}
cachedPreferences = Lists.newArrayListWithCapacity(dataModel.getNumUsers());
cachePreferences();
double rmse = dataModel.getMaxPreference() - dataModel.getMinPreference();
for (int ii = 0; ii < numFeatures; ii++) {
Collections.shuffle(cachedPreferences, random);
for (int i = 0; i < numIterations; i++) {
double err = 0.0;
for (SVDPreference pref : cachedPreferences) {
int useridx = userIndex(pref.getUserID());
int itemidx = itemIndex(pref.getItemID());
err += Math.pow(train(useridx, itemidx, ii, pref), 2.0);
}
rmse = Math.sqrt(err / cachedPreferences.size());
}
if (ii < numFeatures - 1) {
for (SVDPreference pref : cachedPreferences) {
int useridx = userIndex(pref.getUserID());
int itemidx = itemIndex(pref.getItemID());
buildCache(useridx, itemidx, ii, pref);
}
}
log.info("Finished training feature {} with RMSE {}.", ii, rmse);
}
return createFactorization(leftVectors, rightVectors);
}
double getAveragePreference() throws TasteException {
RunningAverage average = new FullRunningAverage();
LongPrimitiveIterator it = dataModel.getUserIDs();
while (it.hasNext()) {
for (Preference pref : dataModel.getPreferencesFromUser(it.nextLong())) {
average.addDatum(pref.getValue());
}
}
return average.getAverage();
}
private double train(int i, int j, int f, SVDPreference pref) {
double[] leftVectorI = leftVectors[i];
double[] rightVectorJ = rightVectors[j];
double prediction = predictRating(i, j, f, pref, true);
double err = pref.getValue() - prediction;
double leftVectorIF = leftVectorI[f];
leftVectorI[f] += learningRate * (err * rightVectorJ[f] - preventOverfitting * leftVectorI[f]);
rightVectorJ[f] += learningRate * (err * leftVectorIF - preventOverfitting * rightVectorJ[f]);
return err;
}
private void buildCache(int i, int j, int k, SVDPreference pref) {
pref.setCache(predictRating(i, j, k, pref, false));
}
private double predictRating(int i, int j, int f, SVDPreference pref, boolean trailing) {
float minPreference = dataModel.getMinPreference();
float maxPreference = dataModel.getMaxPreference();
double sum = pref.getCache();
sum += leftVectors[i][f] * rightVectors[j][f];
if (trailing) {
sum += (numFeatures - f - 1) * (defaultValue + interval) * (defaultValue + interval);
if (sum > maxPreference) {
sum = maxPreference;
} else if (sum < minPreference) {
sum = minPreference;
}
}
return sum;
}
private void cachePreferences() throws TasteException {
cachedPreferences.clear();
LongPrimitiveIterator it = dataModel.getUserIDs();
while (it.hasNext()) {
for (Preference pref : dataModel.getPreferencesFromUser(it.nextLong())) {
cachedPreferences.add(new SVDPreference(pref.getUserID(), pref.getItemID(), pref.getValue(), 0.0));
}
}
}
}