-
Notifications
You must be signed in to change notification settings - Fork 3
/
kmeans_hirschberg_larmore.cpp
216 lines (193 loc) · 5.99 KB
/
kmeans_hirschberg_larmore.cpp
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
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
#include <algorithm>
#include <cassert>
#include <cmath>
#include <cstdint>
#include <cstdlib>
#include <cstring>
#include <deque>
#include <iomanip>
#include <iostream>
#include <limits>
#include <memory>
#include <tuple>
#include <utility>
#include <vector>
#include "kmeans.h"
#include "interval_sum.hpp"
static double lambda = 0;
static interval_sum<double> is;
static std::vector<double> f;
static std::vector<size_t> bestleft;
kmeans_hirschberg_larmore::kmeans_hirschberg_larmore(const std::vector<double> &points) :
f(points.size() + 1, 0.0), bestleft(points.size() + 1, 0),
is(points), points(points), n(points.size()) { }
std::string kmeans_hirschberg_larmore::name() { return std::string("hirc"); }
std::unique_ptr<kmeans_result> kmeans_hirschberg_larmore::compute(size_t k) {
std::unique_ptr<kmeans_result> kmeans_res(new kmeans_result);
if (k >= n) {
kmeans_res->cost = 0.0;
kmeans_res->centers.resize(k);
for (size_t i = 0; i < n; ++i) {
kmeans_res->centers[i] = points[i];
}
for (size_t i = n; i < k; ++i) {
kmeans_res->centers[i] = points[n-1];
}
return kmeans_res;
}
if (k == 1) {
kmeans_res->cost = is.cost_interval_l2(0, n-1);
kmeans_res->centers.push_back(is.query(0, n) / ((double) n));
return kmeans_res;
}
double lo = 0.0;
double hi = is.cost_interval_l2(0, n-1);
double hi_intercept = hi;
double lo_intercept = 0;
size_t hi_k = 1;
size_t lo_k = n;
//double hi = 1e-2;
double val_found, val_found2;
size_t k_found, k_found2;
size_t cnt = 0;
while (true) {
++cnt;
double t = (hi_intercept - lo_intercept) / sqrt(lo_k - hi_k);
double intercept_guess = (hi_intercept + lo_intercept) / 2;
double intersect_hi = (intercept_guess - hi_intercept) / (hi_k - k);
double intersect_lo = (intercept_guess - lo_intercept) / (lo_k - k);
assert(intercept_guess > 0);
assert(intercept_guess <= hi_intercept);
assert(intercept_guess >= lo_intercept);
lambda = (hi_intercept - lo_intercept) / (lo_k - hi_k);
std::tie(val_found, k_found) = this->basic(n);
if (k_found > k) {
lo_k = k_found;
lo = lambda;
lo_intercept = val_found - lo_k * lambda;
} else if (k_found < k) {
hi = lambda;
hi_k = k_found;
hi_intercept = val_found - hi_k * lambda;
} else {
hi = lambda;
break;
}
}
assert(k == k_found);
get_actual_cost(n, kmeans_res);
return kmeans_res;
}
std::unique_ptr<kmeans_result> kmeans_hirschberg_larmore::compute_and_report(size_t k) {
return compute(k);
}
double kmeans_hirschberg_larmore::weight(size_t i, size_t j) {
if (i >= j) return std::numeric_limits<double>::max();
return is.cost_interval_l2(i, j-1) + lambda;
}
double kmeans_hirschberg_larmore::g(size_t i, size_t j) {
return f[i] + weight(i, j);
}
bool kmeans_hirschberg_larmore::bridge(size_t i, size_t j, size_t k, size_t n) {
if (k == n) {
return true;
}
if (g(i, n) <= g(j, n)) {
return true;
}
size_t lo = k;
size_t hi = n;
while (hi - lo >= 2) {
size_t mid = lo + (hi-lo)/2;
double gim = g(i, mid);
double gjm = g(j, mid);
double gkm = g(k, mid);
if (gim <= gjm) {
lo = mid;
if (gkm <= gjm) return true;
} else {
hi = mid;
if (gjm < gkm) return false;
}
}
bool result = (g(k, hi) <= g(j, hi));
return result;
}
std::pair<double, size_t> kmeans_hirschberg_larmore::basic(size_t n) {
std::cout << "call basic lambda=" << lambda << std::endl;
f.resize(n+1, 0);
for (size_t i = 0; i <= n; ++i) f[i] = 0;
bestleft.resize(n+1, 0);
for (size_t i = 0; i <= n; ++i) bestleft[i] = 0;
std::vector<size_t> D = {0};
size_t front = 0;
for (size_t m = 1; m <= n-1; ++m) {
f[m] = g(D[front], m);
bestleft[m] = D[front];
while (front + 1 < D.size() && g(D[front + 1], m+1) <= g(D[front], m+1)) {
++front;
}
if (g(m, n) < g(D[D.size() - 1], n)) {
D.push_back(m);
} else { continue; }
while (front + 2 < D.size() && bridge(D[D.size() - 3], D[D.size() - 2], m, n)) {
std::swap(D[D.size() - 1], D[D.size() - 2]);
D.pop_back();
}
if (front + 2 == D.size() && g(D[D.size() - 1], m+1) <= g(D[D.size() - 2], m+1)) {
++front;
}
}
assert(front + 1 == D.size());
f[n] = g(D[front], n);
bestleft[n] = D[front];
// find length.
size_t m = n;
size_t length = 0;
while (m > 0) {
m = bestleft[m];
++length;
}
return std::make_pair(f[n], length);
}
double kmeans_hirschberg_larmore::get_actual_cost(size_t n, std::unique_ptr<kmeans_result> &res) {
double cost = 0.0;
size_t m = n;
std::vector<double> centers;
while (m != 0) {
size_t prev = bestleft[m];
cost += is.cost_interval_l2(prev, m-1);
double avg = is.query(prev, m) / (m - prev);
centers.push_back(avg);
m = prev;
}
res->centers.resize(centers.size());
for (size_t i = 0; i < centers.size(); ++i) {
res->centers[i] = centers[centers.size() - i - 1];
}
res->cost = cost;
return cost;
}
std::pair<double, size_t> kmeans_hirschberg_larmore::traditional(size_t n) {
f.resize(n, 0);
for (size_t i = 0; i < n; ++i) f[i] = 0;
bestleft.resize(n, 0);
for (size_t i = 0; i < n; ++i) bestleft[i] = 0;
for (size_t m = 1; m < n; ++m) {
f[m] = g(0, m);
bestleft[m] = 0;
for (size_t i = 1; i < m; ++i) {
if (g(i, m) < f[m]) {
f[m] = g(i, m);
bestleft[m] = i;
}
}
}
size_t m = n-1;
size_t length = 0;
while (m > 0) {
m = bestleft[m];
++length;
}
return std::make_pair(f[n-1], length);
}