1 | using System;
|
---|
2 | using System.Collections.Generic;
|
---|
3 | using System.Diagnostics;
|
---|
4 | using System.Linq;
|
---|
5 | using System.Text;
|
---|
6 | using System.Text.RegularExpressions;
|
---|
7 | using HeuristicLab.Common;
|
---|
8 |
|
---|
9 | namespace HeuristicLab.Problems.GrammaticalOptimization {
|
---|
10 | // must find one of numCorrectPhrases*sequenceLen sequences where the quality of a sequence is the length of the subsequence containing only correct _phrases_ (of length phraseLen) and starting at the first position
|
---|
11 | // compared to the RoyalSequence problem this problem is harder because the number of different phrases starting at a position is much larger than the number of symbols (grows exponentially with the phrase-length)
|
---|
12 | // if phraseLen = 1 this is the same as the RoyalSequence problem
|
---|
13 | // parameters
|
---|
14 | // - alphabetSize: number of different symbols (max=26)
|
---|
15 | // - phraseLen: the length of a phrase in number of symbols
|
---|
16 | // - sequenceLen: the number of phrases in the correct subsequence (total sequence length is n * phraseLen
|
---|
17 | // - numCorrectPhrases: the number of correct phrases starting at each position
|
---|
18 | // - phrasesAsSets: switch to determine if the ordering of symbols within a phrase is relevant
|
---|
19 | //
|
---|
20 | // this problem should be hard for GP and easy for MCTS (TD should not have an advantage compared to MCTS)
|
---|
21 | // for phraseLen > 1 this should be harder than RoyalSymbolProblem
|
---|
22 | // when phrases are symbol sets instead of sequences then value-estimation routines should be better (TD)
|
---|
23 | public class RoyalPhraseSequenceProblem : IProblem {
|
---|
24 |
|
---|
25 | private readonly IGrammar grammar;
|
---|
26 | private readonly double correctReward;
|
---|
27 | private readonly double incorrectReward;
|
---|
28 | private readonly int _numCorrectPhrases;
|
---|
29 | private readonly int sequenceLen;
|
---|
30 | private readonly int alphabetSize;
|
---|
31 | private readonly int phraseLen;
|
---|
32 | private readonly bool phrasesAsSets;
|
---|
33 | private readonly SortedSet<string>[] optimalPhrasesForPos;
|
---|
34 |
|
---|
35 | public RoyalPhraseSequenceProblem(Random rand, int alphabetSize, int sequenceLen, int phraseLen = 1, int numCorrectPhrases = 1, double correctReward = 1.0, double incorrectReward = 0.0, bool phrasesAsSets = false) {
|
---|
36 | if (alphabetSize <= 0 || alphabetSize > 26) throw new ArgumentException();
|
---|
37 | if (sequenceLen <= 0) throw new ArgumentException();
|
---|
38 | if (numCorrectPhrases < 1 || numCorrectPhrases > alphabetSize) throw new ArgumentException();
|
---|
39 | if (phraseLen < 1) throw new ArgumentException();
|
---|
40 | if (correctReward <= incorrectReward) throw new ArgumentException();
|
---|
41 |
|
---|
42 | this.alphabetSize = alphabetSize;
|
---|
43 | this.sequenceLen = sequenceLen;
|
---|
44 | this.phraseLen = phraseLen;
|
---|
45 | this._numCorrectPhrases = numCorrectPhrases;
|
---|
46 | this.correctReward = correctReward;
|
---|
47 | this.incorrectReward = incorrectReward;
|
---|
48 | this.phrasesAsSets = phrasesAsSets;
|
---|
49 | var sentenceSymbol = 'S';
|
---|
50 | var terminalSymbols = Enumerable.Range(0, alphabetSize).Select(off => (char)((byte)'a' + off)).ToArray();
|
---|
51 | var nonTerminalSymbols = new char[] { 'S' };
|
---|
52 | var rules = terminalSymbols.Select(t => Tuple.Create('S', t.ToString()))
|
---|
53 | .Concat(terminalSymbols.Select(t => Tuple.Create('S', t + "S")));
|
---|
54 |
|
---|
55 | this.grammar = new Grammar(sentenceSymbol, terminalSymbols, nonTerminalSymbols, rules);
|
---|
56 |
|
---|
57 | this.optimalPhrasesForPos = new SortedSet<string>[sequenceLen];
|
---|
58 | for (int i = 0; i < sequenceLen; i++) {
|
---|
59 | optimalPhrasesForPos[i] = new SortedSet<string>();
|
---|
60 | for (int j = 0; j < numCorrectPhrases; j++) {
|
---|
61 | string phrase = "";
|
---|
62 | do {
|
---|
63 | for (int l = 0; l < phraseLen; l++) {
|
---|
64 | phrase += terminalSymbols.SelectRandom(rand);
|
---|
65 | }
|
---|
66 | phrase = CanonicalPhrase(phrase);
|
---|
67 | } while (optimalPhrasesForPos[i].Contains(phrase)); // don't allow duplicate phrases
|
---|
68 | optimalPhrasesForPos[i].Add(phrase);
|
---|
69 | }
|
---|
70 | }
|
---|
71 |
|
---|
72 | Debug.Assert(Evaluate(BestKnownSolution) / BestKnownQuality(phraseLen * sequenceLen) == 1.0);
|
---|
73 | }
|
---|
74 |
|
---|
75 | public double BestKnownQuality(int maxLen) {
|
---|
76 | return Math.Min(maxLen / phraseLen, sequenceLen) * correctReward; // integer division
|
---|
77 | }
|
---|
78 |
|
---|
79 | public string BestKnownSolution {
|
---|
80 | get {
|
---|
81 | string solution = "";
|
---|
82 | for (int i = 0; i < sequenceLen; i++) {
|
---|
83 | solution += optimalPhrasesForPos[i].First();
|
---|
84 | }
|
---|
85 | return solution;
|
---|
86 | }
|
---|
87 | }
|
---|
88 |
|
---|
89 | public IGrammar Grammar {
|
---|
90 | get { return grammar; }
|
---|
91 | }
|
---|
92 |
|
---|
93 | public double Evaluate(string sentence) {
|
---|
94 | // sentence must contain only terminal symbols, we are not checking if the sentence is syntactically valid here because it would be too slow!
|
---|
95 | Debug.Assert(sentence.Any(c => grammar.IsTerminal(c)));
|
---|
96 | // as long as only correct symbols are found we increase the reward by +1
|
---|
97 | // on the first incorrect symbol we return
|
---|
98 | var reward = 0.0;
|
---|
99 | for (int i = 0; i < Math.Min(sentence.Length / phraseLen, sequenceLen); i++) {
|
---|
100 | var canonicalPhrase = CanonicalPhrase(sentence.Substring(i * phraseLen, phraseLen));
|
---|
101 | if (optimalPhrasesForPos[i].Contains(canonicalPhrase)) {
|
---|
102 | reward += correctReward;
|
---|
103 | } else {
|
---|
104 | // alternatively reduce reward by number of remaining phrases
|
---|
105 | return Math.Max(0.0, reward + incorrectReward * (sentence.Length / phraseLen - i));
|
---|
106 | // stop on first incorrect symbol and return reward
|
---|
107 | //return reward;
|
---|
108 | }
|
---|
109 | }
|
---|
110 | return reward;
|
---|
111 | }
|
---|
112 |
|
---|
113 | private string CanonicalPhrase(string phrase) {
|
---|
114 | if (phrasesAsSets) return string.Join("", phrase.OrderBy(ch => (byte)ch));
|
---|
115 | else return phrase;
|
---|
116 | }
|
---|
117 |
|
---|
118 | public string CanonicalRepresentation(string terminalPhrase) {
|
---|
119 | if (phrasesAsSets) {
|
---|
120 | var phrases = new List<string>();
|
---|
121 | var numPhrases = terminalPhrase.Length / phraseLen;
|
---|
122 | for (int phraseIdx = 0; phraseIdx < numPhrases; phraseIdx++) {
|
---|
123 | var sentenceIdx = phraseIdx * phraseLen;
|
---|
124 | var phrase = terminalPhrase.Substring(sentenceIdx, phraseLen);
|
---|
125 | phrase = CanonicalPhrase(phrase);
|
---|
126 | phrases.Add(phrase);
|
---|
127 | }
|
---|
128 |
|
---|
129 | return string.Join("", phrases);
|
---|
130 | } else
|
---|
131 | return terminalPhrase;
|
---|
132 | }
|
---|
133 | }
|
---|
134 | }
|
---|