Free cookie consent management tool by TermsFeed Policy Generator

source: trunk/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/Hashing/SymbolicExpressionTreeHash.cs @ 17076

Last change on this file since 17076 was 17076, checked in by gkronber, 5 years ago

#2950: made some small changes while reviewing.

File size: 15.2 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2019 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
4 *
5 * This file is part of HeuristicLab.
6 *
7 * HeuristicLab is free software: you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation, either version 3 of the License, or
10 * (at your option) any later version.
11 *
12 * HeuristicLab is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
19 */
20#endregion
21
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
26using static HeuristicLab.Problems.DataAnalysis.Symbolic.SymbolicExpressionHashExtensions;
27
28namespace HeuristicLab.Problems.DataAnalysis.Symbolic {
29  public static class SymbolicExpressionTreeHash {
30    private static readonly Addition add = new Addition();
31    private static readonly Subtraction sub = new Subtraction();
32    private static readonly Multiplication mul = new Multiplication();
33    private static readonly Division div = new Division();
34    private static readonly Logarithm log = new Logarithm();
35    private static readonly Exponential exp = new Exponential();
36    private static readonly Sine sin = new Sine();
37    private static readonly Cosine cos = new Cosine();
38    private static readonly Constant constant = new Constant();
39
40    private static ISymbolicExpressionTreeNode ActualRoot(this ISymbolicExpressionTree tree) => tree.Root.GetSubtree(0).GetSubtree(0);
41
42    #region tree hashing
43    public static ulong[] Hash(this ISymbolicExpressionTree tree, bool simplify = false, bool strict = false) {
44      return tree.ActualRoot().Hash(simplify, strict);
45    }
46
47    public static ulong[] Hash(this ISymbolicExpressionTreeNode node, bool simplify = false, bool strict = false) {
48      ulong hashFunction(byte[] input) => HashUtil.DJBHash(input);
49
50      var hashNodes = simplify ? node.MakeNodes(strict).Simplify(hashFunction) : node.MakeNodes(strict).Sort(hashFunction); // simplify sorts implicitly
51      var hashes = new ulong[hashNodes.Length];
52      for (int i = 0; i < hashes.Length; ++i) {
53        hashes[i] = hashNodes[i].CalculatedHashValue;
54      }
55      return hashes;
56    }
57
58    public static ulong ComputeHash(this ISymbolicExpressionTree tree, bool simplify = false, bool strict = false) {
59      return ComputeHash(tree.ActualRoot(), simplify, strict);
60    }
61
62    public static ulong ComputeHash(this ISymbolicExpressionTreeNode treeNode, bool simplify = false, bool strict = false) {
63      return treeNode.Hash(simplify, strict).Last();
64    }
65
66    public static HashNode<ISymbolicExpressionTreeNode> ToHashNode(this ISymbolicExpressionTreeNode node, bool strict = false) {
67      var symbol = node.Symbol;
68      var name = symbol.Name;
69      if (node is ConstantTreeNode constantNode) {
70        name = strict ? constantNode.Value.ToString() : symbol.Name;
71      } else if (node is VariableTreeNode variableNode) {
72        name = strict ? variableNode.Weight.ToString() + variableNode.VariableName : variableNode.VariableName;
73      }
74      var hash = (ulong)name.GetHashCode();
75      var hashNode = new HashNode<ISymbolicExpressionTreeNode> {
76        Data = node,
77        Arity = node.SubtreeCount,
78        Size = node.SubtreeCount,
79        IsCommutative = node.Symbol is Addition || node.Symbol is Multiplication,
80        Enabled = true,
81        HashValue = hash,
82        CalculatedHashValue = hash
83      };
84      if (symbol is Addition) {
85        hashNode.Simplify = SimplifyAddition;
86      } else if (symbol is Multiplication) {
87        hashNode.Simplify = SimplifyMultiplication;
88      } else if (symbol is Division) {
89        hashNode.Simplify = SimplifyDivision;
90      } else if (symbol is Logarithm || symbol is Exponential || symbol is Sine || symbol is Cosine) {
91        hashNode.Simplify = SimplifyUnaryNode;
92      } else if (symbol is Subtraction) {
93        hashNode.Simplify = SimplifyBinaryNode;
94      }
95      return hashNode;
96    }
97
98    public static HashNode<ISymbolicExpressionTreeNode>[] MakeNodes(this ISymbolicExpressionTree tree, bool strict = false) {
99      return MakeNodes(tree.ActualRoot(), strict);
100    }
101
102    public static HashNode<ISymbolicExpressionTreeNode>[] MakeNodes(this ISymbolicExpressionTreeNode node, bool strict = false) {
103      return node.IterateNodesPostfix().Select(x => x.ToHashNode(strict)).ToArray().UpdateNodeSizes();
104    }
105    #endregion
106
107    #region tree similarity
108    public static double ComputeSimilarity(ISymbolicExpressionTree t1, ISymbolicExpressionTree t2, bool simplify = false, bool strict = false) {
109      return ComputeSimilarity(t1.ActualRoot(), t2.ActualRoot(), simplify, strict);
110    }
111
112    public static double ComputeSimilarity(ISymbolicExpressionTreeNode t1, ISymbolicExpressionTreeNode t2, bool simplify = false, bool strict = false) {
113      var lh = t1.Hash(simplify, strict);
114      var rh = t2.Hash(simplify, strict);
115
116      Array.Sort(lh);
117      Array.Sort(rh);
118
119      return ComputeSimilarity(lh, rh);
120    }
121
122    // requires lhs and rhs to be sorted
123    public static int IntersectCount(this ulong[] lh, ulong[] rh) {
124      int count = 0;
125      for (int i = 0, j = 0; i < lh.Length && j < rh.Length;) {
126        var h1 = lh[i];
127        var h2 = rh[j];
128        if (h1 == h2) {
129          ++count;
130          ++i;
131          ++j;
132        } else if (h1 < h2) {
133          ++i;
134        } else if (h1 > h2) {
135          ++j;
136        }
137      }
138      return count;
139    }
140
141    public static IEnumerable<ulong> Intersect(this ulong[] lh, ulong[] rh) {
142      for (int i = 0, j = 0; i < lh.Length && j < rh.Length;) {
143        var h1 = lh[i];
144        var h2 = rh[j];
145        if (h1 == h2) {
146          yield return h1;
147          ++i;
148          ++j;
149        } else if (h1 < h2) {
150          ++i;
151        } else if (h1 > h2) {
152          ++j;
153        }
154      }
155    }
156
157    // this will only work if lh and rh are sorted
158    public static double ComputeSimilarity(ulong[] lh, ulong[] rh) {
159      return 2d * IntersectCount(lh, rh) / (lh.Length + rh.Length);
160    }
161
162    public static double ComputeAverageSimilarity(IList<ISymbolicExpressionTree> trees, bool simplify = false, bool strict = false) {
163      var total = trees.Count * (trees.Count - 1) / 2;
164      double avg = 0;
165      var hashes = new ulong[trees.Count][];
166      // build hash arrays
167      for (int i = 0; i < trees.Count; ++i) {
168        var nodes = trees[i].MakeNodes(strict);
169        hashes[i] = (simplify ? nodes.Simplify(HashUtil.DJBHash) : nodes.Sort(HashUtil.DJBHash)).Select(x => x.CalculatedHashValue).ToArray();
170        Array.Sort(hashes[i]);
171      }
172      // compute similarity matrix
173      for (int i = 0; i < trees.Count - 1; ++i) {
174        for (int j = i + 1; j < trees.Count; ++j) {
175          avg += ComputeSimilarity(hashes[i], hashes[j]);
176        }
177      }
178      return avg / total;
179    }
180
181    public static double[,] ComputeSimilarityMatrix(IList<ISymbolicExpressionTree> trees, bool simplify = false, bool strict = false) {
182      var sim = new double[trees.Count, trees.Count];
183      var hashes = new ulong[trees.Count][];
184      // build hash arrays
185      for (int i = 0; i < trees.Count; ++i) {
186        var nodes = trees[i].MakeNodes(strict);
187        hashes[i] = (simplify ? nodes.Simplify(HashUtil.DJBHash) : nodes.Sort(HashUtil.DJBHash)).Select(x => x.CalculatedHashValue).ToArray();
188        Array.Sort(hashes[i]);
189      }
190      // compute similarity matrix
191      for (int i = 0; i < trees.Count - 1; ++i) {
192        for (int j = i + 1; j < trees.Count; ++j) {
193          sim[i, j] = sim[j, i] = ComputeSimilarity(hashes[i], hashes[j]);
194        }
195      }
196      return sim;
197    }
198    #endregion
199
200    #region parse a nodes array back into a tree
201    public static ISymbolicExpressionTree ToTree(this HashNode<ISymbolicExpressionTreeNode>[] nodes) {
202      var root = new ProgramRootSymbol().CreateTreeNode();
203      var start = new StartSymbol().CreateTreeNode();
204      root.AddSubtree(start);
205      start.AddSubtree(nodes.ToSubtree());
206      return new SymbolicExpressionTree(root);
207    }
208
209    public static ISymbolicExpressionTreeNode ToSubtree(this HashNode<ISymbolicExpressionTreeNode>[] nodes) {
210      var treeNodes = nodes.Select(x => x.Data.Symbol.CreateTreeNode()).ToArray();
211
212      // construct tree top down (assumes postfix order for nodes)
213      for (int i = nodes.Length - 1; i >= 0; --i) {
214        var node = nodes[i];
215
216        if (node.IsLeaf) {
217          if (node.Data is VariableTreeNode variable) {
218            var variableTreeNode = (VariableTreeNode)treeNodes[i];
219            variableTreeNode.VariableName = variable.VariableName;
220            variableTreeNode.Weight = variable.Weight;
221          } else if (node.Data is ConstantTreeNode @const) {
222            var constantTreeNode = (ConstantTreeNode)treeNodes[i];
223            constantTreeNode.Value = @const.Value;
224          }
225          continue;
226        }
227
228        var treeNode = treeNodes[i];
229
230        foreach (var j in nodes.IterateChildren(i)) {
231          treeNode.AddSubtree(treeNodes[j]);
232        }
233      }
234
235      return treeNodes.Last();
236    }
237
238    private static T CreateTreeNode<T>(this ISymbol symbol) where T : class, ISymbolicExpressionTreeNode {
239      return (T)symbol.CreateTreeNode();
240    }
241    #endregion
242
243    #region tree simplification
244    // these simplification methods rely on the assumption that child nodes of the current node have already been simplified
245    // (in other words simplification should be applied in a bottom-up fashion)
246    public static ISymbolicExpressionTree Simplify(ISymbolicExpressionTree tree) {
247      ulong hashFunction(byte[] bytes) => HashUtil.DJBHash(bytes);
248      var root = tree.ActualRoot();
249      var nodes = root.MakeNodes();
250      var simplified = nodes.Simplify(hashFunction);
251      return simplified.ToTree();
252    }
253
254    public static void SimplifyAddition(ref HashNode<ISymbolicExpressionTreeNode>[] nodes, int i) {
255      // simplify additions of terms by eliminating terms with the same symbol and hash
256      var children = nodes.IterateChildren(i);
257
258      // we always assume the child nodes are sorted
259      var curr = children[0];
260      var node = nodes[i];
261
262      foreach (var j in children.Skip(1)) {
263        if (nodes[j] == nodes[curr]) {
264          nodes.SetEnabled(j, false);
265          node.Arity--;
266        } else {
267          curr = j;
268        }
269      }
270      if (node.Arity == 1) { // if the arity is 1 we don't need the addition node at all
271        node.Enabled = false;
272      }
273    }
274
275    // simplify multiplications by reducing constants and div terms
276    public static void SimplifyMultiplication(ref HashNode<ISymbolicExpressionTreeNode>[] nodes, int i) {
277      var node = nodes[i];
278      var children = nodes.IterateChildren(i);
279
280      for (int j = 0; j < children.Length; ++j) {
281        var c = children[j];
282        var child = nodes[c];
283
284        if (!child.Enabled)
285          continue;
286
287        var symbol = child.Data.Symbol;
288        if (symbol is Constant) {
289          for (int k = j + 1; k < children.Length; ++k) {
290            var d = children[k];
291            if (nodes[d].Data.Symbol is Constant) {
292              nodes[d].Enabled = false;
293              node.Arity--;
294            } else {
295              break;
296            }
297          }
298        } else if (symbol is Division) {
299          var div = nodes[c];
300          var denominator =
301            div.Arity == 1 ?
302            nodes[c - 1] :                    // 1 / x is expressed as div(x) (with a single child)
303            nodes[c - nodes[c - 1].Size - 2]; // assume division always has arity 1 or 2
304
305          foreach (var d in children) {
306            if (nodes[d].Enabled && nodes[d] == denominator) {
307              nodes[c].Enabled = nodes[d].Enabled = denominator.Enabled = false;
308              node.Arity -= 2; // matching child + division node
309              break;
310            }
311          }
312        }
313
314        if (node.Arity == 0) { // if everything is simplified this node becomes constant
315          var constantTreeNode = constant.CreateTreeNode<ConstantTreeNode>();
316          constantTreeNode.Value = 1;
317          nodes[i] = constantTreeNode.ToHashNode();
318        } else if (node.Arity == 1) { // when i have only 1 arg left i can skip this node
319          node.Enabled = false;
320        }
321      }
322    }
323
324    public static void SimplifyDivision(ref HashNode<ISymbolicExpressionTreeNode>[] nodes, int i) {
325      var node = nodes[i];
326      var children = nodes.IterateChildren(i);
327
328      var tmp = nodes;
329
330      if (children.All(x => tmp[x].Data.Symbol is Constant)) {
331        var v = ((ConstantTreeNode)nodes[children.First()].Data).Value;
332        if (node.Arity == 1) {
333          v = 1 / v;
334        } else if (node.Arity > 1) {
335          foreach (var j in children.Skip(1)) {
336            v /= ((ConstantTreeNode)nodes[j].Data).Value;
337          }
338        }
339        var constantTreeNode = constant.CreateTreeNode<ConstantTreeNode>();
340        constantTreeNode.Value = v;
341        nodes[i] = constantTreeNode.ToHashNode();
342        return;
343      }
344
345      var nominator = nodes[children[0]];
346      foreach (var j in children.Skip(1)) {
347        var denominator = nodes[j];
348        if (nominator == denominator) {
349          // disable all the children of the division node (nominator and children + denominator and children)
350          nominator.Enabled = denominator.Enabled = false;
351          node.Arity -= 2; // nominator + denominator
352        }
353        if (node.Arity == 0) {
354          var constantTreeNode = constant.CreateTreeNode<ConstantTreeNode>();
355          constantTreeNode.Value = 1; // x / x = 1
356          nodes[i] = constantTreeNode.ToHashNode();
357        }
358      }
359    }
360
361    public static void SimplifyUnaryNode(ref HashNode<ISymbolicExpressionTreeNode>[] nodes, int i) {
362      // check if the child of the unary node is a constant, then the whole node can be simplified
363      var parent = nodes[i];
364      var child = nodes[i - 1];
365
366      var parentSymbol = parent.Data.Symbol;
367      var childSymbol = child.Data.Symbol;
368
369      if (childSymbol is Constant) {
370        nodes[i].Enabled = false;
371      } else if ((parentSymbol is Exponential && childSymbol is Logarithm) || (parentSymbol is Logarithm && childSymbol is Exponential)) {
372        // exp(log(x)) == x only for positive x. We consider this as equivalent for hashing anyway.
373        child.Enabled = parent.Enabled = false;
374      }
375    }
376
377    public static void SimplifyBinaryNode(ref HashNode<ISymbolicExpressionTreeNode>[] nodes, int i) {
378      var children = nodes.IterateChildren(i);
379      var tmp = nodes;
380      if (children.All(x => tmp[x].Data.Symbol is Constant)) {
381        foreach (var j in children) {
382          nodes[j].Enabled = false;
383        }
384        nodes[i] = constant.CreateTreeNode().ToHashNode();
385      }
386    }
387    #endregion
388  }
389}
Note: See TracBrowser for help on using the repository browser.