1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
|
// Copyright 2016 Yahoo Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
package com.yahoo.tensor;
import com.google.common.collect.ImmutableMap;
import java.util.ArrayList;
import java.util.Collections;
import java.util.List;
import java.util.ListIterator;
import java.util.Map;
import java.util.Set;
/**
* Computes a <i>sparse tensor product</i>, see {@link Tensor#multiply}
*
* @author bratseth
*/
class TensorProduct {
private final Set<String> dimensionsA, dimensionsB;
private final Set<String> dimensions;
private final ImmutableMap.Builder<TensorAddress, Double> cells = new ImmutableMap.Builder<>();
public TensorProduct(Tensor a, Tensor b) {
dimensionsA = a.dimensions();
dimensionsB = b.dimensions();
// Dimension product
dimensions = TensorOperations.combineDimensions(a, b);
// Cell product (slow baseline implementation)
for (Map.Entry<TensorAddress, Double> aCell : a.cells().entrySet()) {
for (Map.Entry<TensorAddress, Double> bCell : b.cells().entrySet()) {
TensorAddress combinedAddress = combine(aCell.getKey(), bCell.getKey());
if (combinedAddress == null) continue; // not combinable
cells.put(combinedAddress, aCell.getValue() * bCell.getValue());
}
}
}
private TensorAddress combine(TensorAddress a, TensorAddress b) {
List<TensorAddress.Element> combined = new ArrayList<>();
combined.addAll(dense(a, dimensionsA));
combined.addAll(dense(b, dimensionsB));
Collections.sort(combined);
TensorAddress.Element previous = null;
for (ListIterator<TensorAddress.Element> i = combined.listIterator(); i.hasNext(); ) {
TensorAddress.Element current = i.next();
if (previous != null && previous.dimension().equals(current.dimension())) { // an overlapping dimension
if (previous.label().equals(current.label()))
i.remove(); // a match: remove the duplicate
else
return null; // no match: a combination isn't viable
}
previous = current;
}
return TensorAddress.fromSorted(sparse(combined));
}
/**
* Returns a set of tensor elements which contains an entry for each dimension including "undefined" values
* (which are not present in the sparse elements list).
*/
private List<TensorAddress.Element> dense(TensorAddress sparse, Set<String> dimensions) {
if (sparse.elements().size() == dimensions.size()) return sparse.elements();
List<TensorAddress.Element> dense = new ArrayList<>(sparse.elements());
for (String dimension : dimensions) {
if ( ! sparse.hasDimension(dimension))
dense.add(new TensorAddress.Element(dimension, TensorAddress.Element.undefinedLabel));
}
return dense;
}
/**
* Removes any "undefined" entries from the given elements.
*/
private List<TensorAddress.Element> sparse(List<TensorAddress.Element> dense) {
List<TensorAddress.Element> sparse = new ArrayList<>();
for (TensorAddress.Element element : dense) {
if ( ! element.label().equals(TensorAddress.Element.undefinedLabel))
sparse.add(element);
}
return sparse;
}
/** Returns the result of taking this product */
public Tensor result() {
return new MapTensor(dimensions, cells.build());
}
}
|