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
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
|
#include <cmath>
#include <iomanip>
#include "MiraFeatureVector.h"
using namespace std;
namespace MosesTuning
{
void MiraFeatureVector::InitSparse(const SparseVector& sparse, size_t ignoreLimit)
{
vector<size_t> sparseFeats = sparse.feats();
bool bFirst = true;
size_t lastFeat = 0;
m_sparseFeats.reserve(sparseFeats.size());
m_sparseVals.reserve(sparseFeats.size());
for(size_t i=0; i<sparseFeats.size(); i++) {
if (sparseFeats[i] < ignoreLimit) continue;
size_t feat = m_dense.size() + sparseFeats[i];
m_sparseFeats.push_back(feat);
m_sparseVals.push_back(sparse.get(sparseFeats[i]));
// Check ordered property
if(bFirst) {
bFirst = false;
} else {
if(lastFeat>=feat) {
cerr << "Error: Feature indeces must be strictly ascending coming out of SparseVector" << endl;
exit(1);
}
}
lastFeat = feat;
}
}
MiraFeatureVector::MiraFeatureVector(const FeatureDataItem& vec)
: m_dense(vec.dense)
{
InitSparse(vec.sparse);
}
MiraFeatureVector::MiraFeatureVector(const SparseVector& sparse, size_t num_dense)
{
m_dense.resize(num_dense);
//Assume that features with id [0,num_dense) are the dense features
for (size_t id = 0; id < num_dense; ++id) {
m_dense[id] = sparse.get(id);
}
InitSparse(sparse,num_dense);
}
MiraFeatureVector::MiraFeatureVector(const MiraFeatureVector& other)
: m_dense(other.m_dense),
m_sparseFeats(other.m_sparseFeats),
m_sparseVals(other.m_sparseVals)
{
if(m_sparseVals.size()!=m_sparseFeats.size()) {
cerr << "Error: mismatching sparse feat and val sizes" << endl;
exit(1);
}
}
MiraFeatureVector::MiraFeatureVector(const vector<ValType>& dense,
const vector<size_t>& sparseFeats,
const vector<ValType>& sparseVals)
: m_dense(dense),
m_sparseFeats(sparseFeats),
m_sparseVals(sparseVals)
{
if(m_sparseVals.size()!=m_sparseFeats.size()) {
cerr << "Error: mismatching sparse feat and val sizes" << endl;
exit(1);
}
}
ValType MiraFeatureVector::val(size_t index) const
{
if(index < m_dense.size())
return m_dense[index];
else
return m_sparseVals[index-m_dense.size()];
}
size_t MiraFeatureVector::feat(size_t index) const
{
if(index < m_dense.size())
return index;
else
return m_sparseFeats[index-m_dense.size()];
}
size_t MiraFeatureVector::size() const
{
return m_dense.size() + m_sparseVals.size();
}
ValType MiraFeatureVector::sqrNorm() const
{
ValType toRet = 0.0;
for(size_t i=0; i<m_dense.size(); i++)
toRet += m_dense[i]*m_dense[i];
for(size_t i=0; i<m_sparseVals.size(); i++)
toRet += m_sparseVals[i] * m_sparseVals[i];
return toRet;
}
MiraFeatureVector operator-(const MiraFeatureVector& a, const MiraFeatureVector& b)
{
// Dense subtraction
vector<ValType> dense;
if(a.m_dense.size()!=b.m_dense.size()) {
cerr << "Mismatching dense vectors passed to MiraFeatureVector subtraction" << endl;
exit(1);
}
for(size_t i=0; i<a.m_dense.size(); i++) {
dense.push_back(a.m_dense[i] - b.m_dense[i]);
}
// Sparse subtraction
size_t i=0;
size_t j=0;
vector<ValType> sparseVals;
vector<size_t> sparseFeats;
while(i < a.m_sparseFeats.size() && j < b.m_sparseFeats.size()) {
if(a.m_sparseFeats[i] < b.m_sparseFeats[j]) {
sparseFeats.push_back(a.m_sparseFeats[i]);
sparseVals.push_back(a.m_sparseVals[i]);
i++;
}
else if(b.m_sparseFeats[j] < a.m_sparseFeats[i]) {
sparseFeats.push_back(b.m_sparseFeats[j]);
sparseVals.push_back(-b.m_sparseVals[j]);
j++;
}
else {
ValType newVal = a.m_sparseVals[i] - b.m_sparseVals[j];
if(abs(newVal)>1e-6) {
sparseFeats.push_back(a.m_sparseFeats[i]);
sparseVals.push_back(newVal);
}
i++;
j++;
}
}
while(i<a.m_sparseFeats.size()) {
sparseFeats.push_back(a.m_sparseFeats[i]);
sparseVals.push_back(a.m_sparseVals[i]);
i++;
}
while(j<b.m_sparseFeats.size()) {
sparseFeats.push_back(b.m_sparseFeats[j]);
sparseVals.push_back(-b.m_sparseVals[j]);
j++;
}
// Create and return vector
return MiraFeatureVector(dense,sparseFeats,sparseVals);
}
bool operator==(const MiraFeatureVector& a,const MiraFeatureVector& b)
{
ValType eps = 1e-8;
//dense features
if (a.m_dense.size() != b.m_dense.size()) return false;
for (size_t i = 0; i < a.m_dense.size(); ++i) {
if (fabs(a.m_dense[i]-b.m_dense[i]) < eps) return false;
}
if (a.m_sparseFeats.size() != b.m_sparseFeats.size()) return false;
for (size_t i = 0; i < a.m_sparseFeats.size(); ++i) {
if (a.m_sparseFeats[i] != b.m_sparseFeats[i]) return false;
if (fabs(a.m_sparseVals[i] != b.m_sparseVals[i])) return false;
}
return true;
}
ostream& operator<<(ostream& o, const MiraFeatureVector& e)
{
for(size_t i=0; i<e.size(); i++) {
if(i>0) o << " ";
o << e.feat(i) << ":" << e.val(i);
}
return o;
}
// --Emacs trickery--
// Local Variables:
// mode:c++
// c-basic-offset:2
// End:
}
|
