001/* 002 * Copyright (c) 2015-2020, Oracle and/or its affiliates. All rights reserved. 003 * 004 * Licensed under the Apache License, Version 2.0 (the "License"); 005 * you may not use this file except in compliance with the License. 006 * You may obtain a copy of the License at 007 * 008 * http://www.apache.org/licenses/LICENSE-2.0 009 * 010 * Unless required by applicable law or agreed to in writing, software 011 * distributed under the License is distributed on an "AS IS" BASIS, 012 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express implied. 013 * See the License for the specific language governing permissions and 014 * limitations under the License. 015 */ 016 017package org.tribuo.regression.baseline; 018 019import com.oracle.labs.mlrg.olcut.util.Pair; 020import org.tribuo.Example; 021import org.tribuo.Excuse; 022import org.tribuo.ImmutableFeatureMap; 023import org.tribuo.ImmutableOutputInfo; 024import org.tribuo.Model; 025import org.tribuo.Prediction; 026import org.tribuo.Trainer; 027import org.tribuo.provenance.ModelProvenance; 028import org.tribuo.regression.Regressor; 029import org.tribuo.regression.baseline.DummyRegressionTrainer.DummyType; 030 031import java.util.Arrays; 032import java.util.Collections; 033import java.util.List; 034import java.util.Map; 035import java.util.Optional; 036import java.util.Random; 037 038/** 039 * A model which performs dummy regressions (e.g., constant output, gaussian sampled output, mean value, median, quartile). 040 */ 041public class DummyRegressionModel extends Model<Regressor> { 042 private static final long serialVersionUID = 2L; 043 044 private final DummyType dummyType; 045 046 private final Regressor output; 047 048 private final long seed; 049 050 private final Random rng; 051 052 private final double[] means; 053 054 private final double[] variances; 055 056 private final String[] dimensionNames; 057 058 DummyRegressionModel(ModelProvenance description, ImmutableFeatureMap featureIDMap, ImmutableOutputInfo<Regressor> outputIDInfo, long seed, double[] means, double[] variances, String[] names) { 059 super("dummy-GAUSSIAN-regression", description, featureIDMap, outputIDInfo, false); 060 this.dummyType = DummyType.GAUSSIAN; 061 this.output = null; 062 this.seed = seed; 063 this.rng = new Random(seed); 064 this.means = Arrays.copyOf(means,means.length); 065 this.variances = Arrays.copyOf(variances,variances.length); 066 this.dimensionNames = Arrays.copyOf(names,names.length); 067 } 068 069 DummyRegressionModel(ModelProvenance description, ImmutableFeatureMap featureIDMap, ImmutableOutputInfo<Regressor> outputIDInfo, DummyType dummyType, Regressor regressor) { 070 super("dummy-"+dummyType+"-regression", description, featureIDMap, outputIDInfo, false); 071 this.dummyType = dummyType; 072 this.output = regressor; 073 this.seed = Trainer.DEFAULT_SEED; 074 this.rng = null; 075 this.means = new double[0]; 076 this.variances = new double[0]; 077 this.dimensionNames = new String[0]; 078 } 079 080 @Override 081 public Prediction<Regressor> predict(Example<Regressor> example) { 082 switch (dummyType) { 083 case CONSTANT: 084 case MEAN: 085 case MEDIAN: 086 case QUARTILE: 087 return new Prediction<>(output,0,example); 088 case GAUSSIAN: { 089 Regressor.DimensionTuple[] dimensions = new Regressor.DimensionTuple[dimensionNames.length]; 090 for (int i = 0; i < dimensionNames.length; i++) { 091 double regressionValue = (rng.nextGaussian() * variances[i]) + means[i]; 092 dimensions[i] = new Regressor.DimensionTuple(dimensionNames[i],regressionValue); 093 } 094 return new Prediction<>(new Regressor(dimensions), 0, example); 095 } 096 default: 097 throw new IllegalStateException("Unknown dummyType " + dummyType); 098 } 099 } 100 101 @Override 102 public Map<String, List<Pair<String, Double>>> getTopFeatures(int n) { 103 if (n != 0) { 104 return Collections.singletonMap(Model.ALL_OUTPUTS, Collections.singletonList(new Pair<>(BIAS_FEATURE, 1.0))); 105 } else { 106 return Collections.emptyMap(); 107 } 108 } 109 110 @Override 111 public Optional<Excuse<Regressor>> getExcuse(Example<Regressor> example) { 112 return Optional.of(new Excuse<>(example,predict(example),getTopFeatures(1))); 113 } 114 115 @Override 116 protected Model<Regressor> copy(String newName, ModelProvenance newProvenance) { 117 switch (dummyType) { 118 case GAUSSIAN: 119 return new DummyRegressionModel(newProvenance,featureIDMap,outputIDInfo,seed,means,variances,dimensionNames); 120 case CONSTANT: 121 case MEAN: 122 case MEDIAN: 123 case QUARTILE: 124 return new DummyRegressionModel(newProvenance,featureIDMap,outputIDInfo,dummyType,output.copy()); 125 default: 126 throw new IllegalStateException("Unknown dummyType " + dummyType); 127 } 128 } 129}