-
Notifications
You must be signed in to change notification settings - Fork 6
Expand file tree
/
Copy pathmain.cpp
More file actions
403 lines (365 loc) · 11.2 KB
/
main.cpp
File metadata and controls
403 lines (365 loc) · 11.2 KB
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
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
// This file is part of SmallBASIC
//
// Plugin for mlpack library - http://mlpack.org
//
// This program is distributed under the terms of the GPL v2.0 or later
// Download the GNU Public License (GPL) from www.gnu.org
//
// Copyright(C) 2023 Chris Warren-Smith
#include "config.h"
#include <sys/stat.h>
#include <cstring>
#include <cstdlib>
#include <cstdio>
#include "include/param.h"
#include "robin-hood-hashing/src/include/robin_hood.h"
#include <mlpack/methods/kmeans/kmeans.hpp>
#include <mlpack/src/mlpack/methods/ann.hpp>
//
// References:
// https://arma.sourceforge.net/docs.html
// https://training.galaxyproject.org/training-material/topics/statistics/tutorials/FNN/tutorial.html
//
// NegativeLogLikelihood is the output layer that is used for classification problem.
// RandomInitialization means that initial weights in neurons are generated randomly in the interval from -1 to 1.
// For regression problems, we use Mean Squared Error (MSE) loss function, which averages the square of the
// difference between predicted and actual values for the batch.
using namespace mlpack;
typedef FFN<MeanSquaredError, GaussianInitialization> FFN_MSE;
typedef FFN<NegativeLogLikelihood, RandomInitialization> FFN_NLL;
typedef FFN<ReconstructionLoss, HeInitialization> FFN_RL;
robin_hood::unordered_map<int, arma::mat *> _dataMap;
robin_hood::unordered_map<int, FFN_MSE *> _ffnMseMap;
robin_hood::unordered_map<int, FFN_NLL *> _ffnNllMap;
robin_hood::unordered_map<int, FFN_RL *> _ffnRlMap;
int _nextId = 1;
#define CLS_DATA 1
#define CLS_FFN_MSE 2
#define CLS_FFN_NLL 3
#define CLS_FFN_RL 4
static void v_setmat(var_t *var, arma::mat *mat) {
int id = ++_nextId;
_dataMap[id] = mat;
map_init_id(var, id, CLS_DATA);
v_setint(map_add_var(var, "n_rows", 0), mat->n_rows);
v_setint(map_add_var(var, "n_cols", 0), mat->n_cols);
}
static int get_data_id(int argc, slib_par_t *params, int arg, var_t *retval) {
int result = -1;
if (is_param_map(argc, params, arg)) {
int id = get_id(params, arg);
if (id != -1 && _dataMap.find(id) != _dataMap.end()) {
result = id;
} else {
error(retval, "Data ID not found");
}
} else {
error(retval, "Map type not found");
}
return result;
}
//
// load from CSV
//
static bool get_mat(arma::mat &dataset, const char *str) {
return data::Load(str, dataset, false);
}
//
// load dataset from SB array
//
static bool get_mat(arma::mat &dataset, var_t *var) {
bool result;
if (v_maxdim(var) == 1) {
// array of numbers or arrays
size_t rows = v_asize(var);
size_t cols = 1;
for (size_t r = 0; r < rows; r++) {
var_t *elem = v_elem(var, r);
if (v_is_type(elem, V_ARRAY)) {
size_t num_elements = v_asize(elem);
if (num_elements > cols) {
cols = num_elements;
}
}
}
dataset.zeros(rows, cols);
for (size_t r = 0; r < rows; r++) {
var_t *elem = v_elem(var, r);
if (v_is_type(elem, V_ARRAY)) {
for (size_t c = 0; c < v_asize(elem); c++) {
dataset(r, c) = get_num(v_elem(elem, c));
}
} else {
dataset(r, 0) = get_num(elem);
}
}
result = true;
} else if (v_maxdim(var) == 2) {
// 2d array
size_t rows = abs(v_ubound(var, 0) - v_lbound(var, 0)) + 1;
size_t cols = abs(v_ubound(var, 1) - v_lbound(var, 1)) + 1;
dataset.zeros(rows, cols);
for (size_t r = 0; r < rows; r++) {
for (size_t c = 0; c < cols; c++) {
size_t pos = r * cols + c;
dataset(r, c) = get_num(v_elem(var, pos));
}
}
result = true;
} else {
result = false;
}
return result;
}
static void neighbor_search(arma::mat &dataset, var_t *retval) {
// dataset.raw_print();
NeighborSearch<NearestNeighborSort, ManhattanDistance> nn(dataset);
arma::Mat<size_t> neighbors;
arma::mat distances;
nn.Search(1, neighbors, distances);
map_init(retval);
var_t *v_neighbors = map_add_var(retval, "neighbors", 0);
var_t *v_distances = map_add_var(retval, "distances", 0);
v_toarray1(v_neighbors, neighbors.n_elem);
v_toarray1(v_distances, neighbors.n_elem);
for (size_t c = 0; c < neighbors.n_elem; c++) {
v_setreal(v_elem(v_neighbors, c), neighbors[c]);
v_setreal(v_elem(v_distances, c), distances[c]);
}
}
static void kmeans_cluster(arma::mat &dataset, size_t clusters, size_t maxIterations, var_t *retval) {
// Create the k-means object and set the parameters.
KMeans<> k(maxIterations);
// The assignments will be stored in this vector.
arma::Row<size_t> assignments;
// The centroids will be stored in this matrix.
arma::mat centroids;
k.Cluster(dataset, clusters, assignments, centroids);
map_init(retval);
var_t *vector = map_add_var(retval, "assignments", 0);
var_t *array = map_add_var(retval, "centroids", 0);
v_toarray1(vector, assignments.n_cols);
for (size_t c = 0; c < assignments.n_cols; c++) {
v_setreal(v_elem(vector, c), assignments(0, c));
}
v_tomatrix(array, centroids.n_rows, centroids.n_cols);
for (size_t r = 0; r < centroids.n_rows; r++) {
for (size_t c = 0; c < centroids.n_cols; c++) {
size_t pos = r * centroids.n_cols + c;
v_setreal(v_elem(array, pos), centroids(r, c));
}
}
}
//
// The assignments vector indicates which cluster each data point
// belongs to (the indices of the vector correspond to the rows of the
// input data matrix), and the centroids matrix contains the
// coordinates of the cluster centroids.
//
// Scheme for clustering image data:
//
// Input - image bytes flattened to a '1D' array (2D for RGBA values)
// r|--------|
// g|--------|
// b|--------|
// a|--------|
//
// Output:
// 'centroids' vector indexes centroids to assignments
// |00001100|
//
// 'assignments' matrix, holding a column position for centroid 0 or 1
// r|n|n|
// g|n|n|
// b|n|n|
// a|n|n|
//
static int cmd_kmeans_cluster(int argc, slib_par_t *params, var_t *retval) {
int result;
int data_id = get_data_id(argc, params, 0, retval);
if (data_id != -1) {
size_t clusters = get_param_int(argc, params, 1, 5);
size_t maxIterations = get_param_int(argc, params, 2, 1000);
kmeans_cluster(*_dataMap.at(data_id), clusters, maxIterations, retval);
result = 1;
} else {
result = 0;
}
return result;
}
static int cmd_neighbor_search(int argc, slib_par_t *params, var_t *retval) {
int result;
int data_id = get_data_id(argc, params, 0, retval);
if (data_id != -1) {
neighbor_search(*_dataMap[data_id], retval);
result = 1;
} else {
result = 0;
}
return result;
}
static int cmd_load(int argc, slib_par_t *params, var_t *retval) {
int result;
arma::mat *dataset = new arma::mat();
if (is_param_array(argc, params, 0)) {
result = get_mat(*dataset, params[0].var_p) ? 1 : 0;
} else if (is_param_str(argc, params, 0)) {
result = get_mat(*dataset, get_param_str(argc, params, 0, nullptr)) ? 1 : 0;
} else {
result = 0;
}
if (result) {
v_setmat(retval, dataset);
} else {
delete dataset;
}
return result;
}
static int cmd_submat(int argc, slib_par_t *params, var_t *retval) {
int result;
int data_id = get_data_id(argc, params, 0, retval);
if (data_id != -1) {
auto first_row = get_param_int(argc, params, 1, 0);
auto first_col = get_param_int(argc, params, 2, 0);
auto last_row = get_param_int(argc, params, 3, 0);
auto last_col = get_param_int(argc, params, 4, 0);
auto mat = _dataMap.at(data_id);
arma::mat submat = mat->submat(first_row, first_col, last_row, last_col);
v_setmat(retval, new arma::mat(submat));
result = 1;
} else {
result = 0;
}
return result;
}
static int cmd_split(int argc, slib_par_t *params, var_t *retval) {
int result;
int data_id = get_data_id(argc, params, 0, retval);
if (data_id != -1) {
auto ratio = get_param_num(argc, params, 1, 0);
arma::mat *train = new arma::mat();
arma::mat *test = new arma::mat();
data::Split(*_dataMap[data_id], *train, *test, ratio);
map_init(retval);
var_t *v_train = map_add_var(retval, "train", 0);
var_t *v_test = map_add_var(retval, "test", 0);
v_setmat(v_train, train);
v_setmat(v_test, test);
result = 1;
} else {
result = 0;
}
return result;
}
static int cmd_create_fnn_mse(int argc, slib_par_t *params, var_t *retval) {
int id = ++_nextId;
_ffnMseMap[id] = new FFN_MSE();
map_init_id(retval, id, CLS_FFN_MSE);
return 1;
}
static int cmd_create_fnn_nll(int argc, slib_par_t *params, var_t *retval) {
int id = ++_nextId;
_ffnNllMap[id] = new FFN_NLL();
map_init_id(retval, id, CLS_FFN_NLL);
return 1;
}
static int cmd_create_fnn_rl(int argc, slib_par_t *params, var_t *retval) {
int id = ++_nextId;
_ffnRlMap[id] = new FFN_RL();
map_init_id(retval, id, CLS_FFN_RL);
return 1;
}
static int cmd_get_row(int argc, slib_par_t *params, var_t *retval) {
int result;
int data_id = get_data_id(argc, params, 0, retval);
if (data_id != -1) {
auto index = get_param_int(argc, params, 1, 0);
arma::mat row = _dataMap[data_id]->row(index);
v_setmat(retval, new arma::mat(row));
result = 1;
} else {
result = 0;
}
return result;
}
static int cmd_scale(int argc, slib_par_t *params, var_t *retval) {
int result;
int train_id = get_data_id(argc, params, 0, retval);
int valid_id = get_data_id(argc, params, 1, retval);
if (train_id != -1 && valid_id != -1) {
arma::mat *train = _dataMap[train_id];
arma::mat *valid = _dataMap[valid_id];
data::MinMaxScaler scale;
scale.Fit(*train);
scale.Transform(*train, *train);
scale.Transform(*valid, *valid);
result = 1;
} else {
result = 0;
}
return result;
}
FUNC_SIG lib_func[] = {
{0, 0, "FFN_MSE", cmd_create_fnn_mse},
{0, 0, "FFN_NLL", cmd_create_fnn_nll},
{0, 0, "FFN_RL", cmd_create_fnn_rl},
{1, 1, "LOAD", cmd_load},
{1, 1, "NEIGHBORSEARCH", cmd_neighbor_search},
{2, 2, "GET_ROW", cmd_get_row},
{2, 2, "SPLIT", cmd_split},
{2, 3, "KMEANSCLUSTER", cmd_kmeans_cluster},
{5, 5, "SUBMAT", cmd_submat}
};
FUNC_SIG lib_proc[] = {
{2, 2, "SCALE", cmd_scale}
};
SBLIB_API int sblib_proc_count() {
return (sizeof(lib_proc) / sizeof(lib_proc[0]));
}
SBLIB_API int sblib_func_count() {
return (sizeof(lib_func) / sizeof(lib_func[0]));
}
SBLIB_API void sblib_free(int cls_id, int id) {
if (id != -1) {
switch (cls_id) {
case CLS_DATA:
if (_dataMap.find(id) != _dataMap.end()) {
delete _dataMap.at(id);
_dataMap.erase(id);
}
break;
case CLS_FFN_MSE:
if (_ffnMseMap.find(id) != _ffnMseMap.end()) {
delete _ffnMseMap.at(id);
_ffnMseMap.erase(id);
}
break;
case CLS_FFN_NLL:
if (_ffnNllMap.find(id) != _ffnNllMap.end()) {
delete _ffnNllMap.at(id);
_ffnNllMap.erase(id);
}
break;
case CLS_FFN_RL:
if (_ffnRlMap.find(id) != _ffnRlMap.end()) {
delete _ffnRlMap.at(id);
_ffnRlMap.erase(id);
}
break;
}
}
}
SBLIB_API void sblib_close(void) {
if (!_dataMap.empty()) {
fprintf(stderr, "Matrix leak detected\n");
}
if (!_ffnMseMap.empty()) {
fprintf(stderr, "FFN MSE leak detected\n");
}
if (!_ffnNllMap.empty()) {
fprintf(stderr, "FFN NLL leak detected\n");
}
if (!_ffnRlMap.empty()) {
fprintf(stderr, "FFN RL leak detected\n");
}
}