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
|
#include "projectorGHSOM.h"
#include <mymaths.h>
#include <QDebug>
#include <GHSOM/globals.h>
#include <map>
using namespace std;
ProjectorGHSOM::ProjectorGHSOM()
{
Globals::EXPAND_CYCLES = 10; // after how many iterations through the data we check if we want to expand
Globals::MAX_CYCLES = 0; // maximum number of iterations through data (0: ignored)
Globals::TAU_1 = 0.01; // percentage of remaining error that needs to be explained by each map. 1.0 == standard SOM
Globals::TAU_2 = 1.0; // granularity of the last layer of maps 1.0 == non-hierarchical SOM (1 map only)
Globals::INITIAL_LEARNRATE = 0.7; // how strongly the winner and neighboring units are adapted initially, decreases over time
Globals::NR = 0.0009; // Neighborhood radius
Globals::randomSeed = 1; // change ad libitum
Globals::normInputVectors=NORM_LENGTH; // NONE, LENGTH, INTERVAL. Normalize data by their length or clamp in a 0-1 interval
Globals::INITIAL_X_SIZE = 2; // initial size in the x direction
Globals::INITIAL_Y_SIZE = 2; // initial size in the y direction
Globals::LABELS_NUM = 0; // max number of labels per unit. 0: no labels
}
void ProjectorGHSOM::SetParams(float tau1, float tau2, int xSize, int ySize, int expandCycles, int normalizationType, float learningRate, float neighborhoodRadius)
{
Globals::TAU_1 = tau1;
Globals::TAU_2 = tau2;
Globals::INITIAL_X_SIZE = xSize;
Globals::INITIAL_Y_SIZE = ySize;
Globals::EXPAND_CYCLES = expandCycles;
Globals::normInputVectors = normalizationType;
Globals::INITIAL_LEARNRATE = learningRate;
Globals::NR = neighborhoodRadius;
}
void ProjectorGHSOM::Train(std::vector< fvec > samples, ivec labels)
{
if(!samples.size()) return;
source = samples;
projected = samples;
dim = samples[0].size();
Data_Vector *res = new Data_Vector();
if(Globals::layers) Globals::layers->removeAllElements();
Globals::vectorlength = dim;
Globals::numofallvecs = samples.size();
//Globals::currentDataLoader = new DataLoader();
FOR(i, samples.size())
{
float *data = new float[dim];
FOR(d, dim) data[d] = samples[i][d];
if (Globals::normInputVectors==NORM_LENGTH)
{
data = Globals::normVec(data);
}
char *id = new char[150];
sprintf(id,"sample%d", i+1);
DataItem *di = new DataItem(id,data,dim);
res->addElement(di);
delete [] data;
delete [] id;
}
if (Globals::normInputVectors==NORM_INTERVAL) res = Globals::normIntervalVector(res);
float *mean = Globals::meanVector(res, dim);
qDebug() << "mean" << mean[0] << mean[1];
Globals::dataItems = res;
char **vecDescr = new char *[dim];
FOR(d, dim)
{
vecDescr[d] = new char[150];
sprintf(vecDescr[d], "dim%d", d+1);
}
Globals::vectorDescription = vecDescr;
Globals::initHFM();
Globals::trainHFM();
// Globals::savePath = "/Users/basilio/Desktop";
// Globals::saveHFMAs(HTML);
// if (Globals::SAVE_AS_HTML) Globals::saveHFMAs(HTML);
// if(Globals::SAVE_AS_SOMLIB) Globals::saveHFMAs(SOMLIB);
}
fvec ProjectorGHSOM::Project(const fvec &sample)
{
return sample;
}
|
