1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 2002-2019 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
4 | * and the BEACON Center for the Study of Evolution in Action.
|
---|
5 | *
|
---|
6 | * This file is part of HeuristicLab.
|
---|
7 | *
|
---|
8 | * HeuristicLab is free software: you can redistribute it and/or modify
|
---|
9 | * it under the terms of the GNU General Public License as published by
|
---|
10 | * the Free Software Foundation, either version 3 of the License, or
|
---|
11 | * (at your option) any later version.
|
---|
12 | *
|
---|
13 | * HeuristicLab is distributed in the hope that it will be useful,
|
---|
14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
16 | * GNU General Public License for more details.
|
---|
17 | *
|
---|
18 | * You should have received a copy of the GNU General Public License
|
---|
19 | * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
|
---|
20 | */
|
---|
21 | #endregion
|
---|
22 |
|
---|
23 | using System;
|
---|
24 | using System.Collections.Generic;
|
---|
25 | using System.Linq;
|
---|
26 | using HeuristicLab.Common;
|
---|
27 | using HeuristicLab.Core;
|
---|
28 | using HEAL.Attic;
|
---|
29 | using HeuristicLab.Problems.DataAnalysis;
|
---|
30 |
|
---|
31 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
32 | [StorableType("1BF7BEFB-6739-48AA-89BC-B632E72D148C")]
|
---|
33 | // this class is used as a surrogate for persistence of an actual GBT model
|
---|
34 | // since the actual GBT model would be very large when persisted we only store all necessary information to
|
---|
35 | // recalculate the actual GBT model on demand
|
---|
36 | [Item("Gradient boosted tree model", "")]
|
---|
37 | public sealed class GradientBoostedTreesModelSurrogate : RegressionModel, IGradientBoostedTreesModel {
|
---|
38 | // don't store the actual model!
|
---|
39 | // the actual model is only recalculated when necessary
|
---|
40 | private readonly Lazy<IGradientBoostedTreesModel> actualModel;
|
---|
41 | private IGradientBoostedTreesModel ActualModel {
|
---|
42 | get { return actualModel.Value; }
|
---|
43 | }
|
---|
44 |
|
---|
45 | [Storable]
|
---|
46 | private readonly IRegressionProblemData trainingProblemData;
|
---|
47 | [Storable]
|
---|
48 | private readonly uint seed;
|
---|
49 | [Storable]
|
---|
50 | private ILossFunction lossFunction;
|
---|
51 | [Storable]
|
---|
52 | private double r;
|
---|
53 | [Storable]
|
---|
54 | private double m;
|
---|
55 | [Storable]
|
---|
56 | private double nu;
|
---|
57 | [Storable]
|
---|
58 | private int iterations;
|
---|
59 | [Storable]
|
---|
60 | private int maxSize;
|
---|
61 |
|
---|
62 |
|
---|
63 | public override IEnumerable<string> VariablesUsedForPrediction {
|
---|
64 | get {
|
---|
65 | return ActualModel.Models.SelectMany(x => x.VariablesUsedForPrediction).Distinct().OrderBy(x => x);
|
---|
66 | }
|
---|
67 | }
|
---|
68 |
|
---|
69 | [StorableConstructor]
|
---|
70 | private GradientBoostedTreesModelSurrogate(StorableConstructorFlag _) : base(_) {
|
---|
71 | actualModel = new Lazy<IGradientBoostedTreesModel>(() => RecalculateModel());
|
---|
72 | }
|
---|
73 |
|
---|
74 | private GradientBoostedTreesModelSurrogate(GradientBoostedTreesModelSurrogate original, Cloner cloner)
|
---|
75 | : base(original, cloner) {
|
---|
76 | IGradientBoostedTreesModel clonedModel = null;
|
---|
77 | if (original.ActualModel != null) clonedModel = cloner.Clone(original.ActualModel);
|
---|
78 | actualModel = new Lazy<IGradientBoostedTreesModel>(CreateLazyInitFunc(clonedModel)); // only capture clonedModel in the closure
|
---|
79 |
|
---|
80 | this.trainingProblemData = cloner.Clone(original.trainingProblemData);
|
---|
81 | this.lossFunction = cloner.Clone(original.lossFunction);
|
---|
82 | this.seed = original.seed;
|
---|
83 | this.iterations = original.iterations;
|
---|
84 | this.maxSize = original.maxSize;
|
---|
85 | this.r = original.r;
|
---|
86 | this.m = original.m;
|
---|
87 | this.nu = original.nu;
|
---|
88 | }
|
---|
89 |
|
---|
90 | private Func<IGradientBoostedTreesModel> CreateLazyInitFunc(IGradientBoostedTreesModel clonedModel) {
|
---|
91 | return () => {
|
---|
92 | return clonedModel == null ? RecalculateModel() : clonedModel;
|
---|
93 | };
|
---|
94 | }
|
---|
95 |
|
---|
96 | // create only the surrogate model without an actual model
|
---|
97 | public GradientBoostedTreesModelSurrogate(IRegressionProblemData trainingProblemData, uint seed,
|
---|
98 | ILossFunction lossFunction, int iterations, int maxSize, double r, double m, double nu)
|
---|
99 | : base(trainingProblemData.TargetVariable, "Gradient boosted tree model", string.Empty) {
|
---|
100 | this.trainingProblemData = trainingProblemData;
|
---|
101 | this.seed = seed;
|
---|
102 | this.lossFunction = lossFunction;
|
---|
103 | this.iterations = iterations;
|
---|
104 | this.maxSize = maxSize;
|
---|
105 | this.r = r;
|
---|
106 | this.m = m;
|
---|
107 | this.nu = nu;
|
---|
108 | }
|
---|
109 |
|
---|
110 | // wrap an actual model in a surrograte
|
---|
111 | public GradientBoostedTreesModelSurrogate(IRegressionProblemData trainingProblemData, uint seed,
|
---|
112 | ILossFunction lossFunction, int iterations, int maxSize, double r, double m, double nu,
|
---|
113 | IGradientBoostedTreesModel model)
|
---|
114 | : this(trainingProblemData, seed, lossFunction, iterations, maxSize, r, m, nu) {
|
---|
115 | actualModel = new Lazy<IGradientBoostedTreesModel>(() => model);
|
---|
116 | }
|
---|
117 |
|
---|
118 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
119 | return new GradientBoostedTreesModelSurrogate(this, cloner);
|
---|
120 | }
|
---|
121 |
|
---|
122 | // forward message to actual model (recalculate model first if necessary)
|
---|
123 | public override IEnumerable<double> GetEstimatedValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
124 | return ActualModel.GetEstimatedValues(dataset, rows);
|
---|
125 | }
|
---|
126 |
|
---|
127 | public override IRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData) {
|
---|
128 | return new RegressionSolution(this, (IRegressionProblemData)problemData.Clone());
|
---|
129 | }
|
---|
130 |
|
---|
131 | private IGradientBoostedTreesModel RecalculateModel() {
|
---|
132 | return GradientBoostedTreesAlgorithmStatic.TrainGbm(trainingProblemData, lossFunction, maxSize, nu, r, m, iterations, seed).Model;
|
---|
133 | }
|
---|
134 |
|
---|
135 | public IEnumerable<IRegressionModel> Models {
|
---|
136 | get {
|
---|
137 | return ActualModel.Models;
|
---|
138 | }
|
---|
139 | }
|
---|
140 |
|
---|
141 | public IEnumerable<double> Weights {
|
---|
142 | get {
|
---|
143 | return ActualModel.Weights;
|
---|
144 | }
|
---|
145 | }
|
---|
146 | }
|
---|
147 | }
|
---|