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
|
/*********************************************************************
MLDemos: A User-Friendly visualization toolkit for machine learning
Copyright (C) 2010 Basilio Noris
Contact: mldemos@b4silio.com
This library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
This library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Library General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with this library; if not, write to the Free
Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*********************************************************************/
#include "public.h"
#include "basicMath.h"
#include <drawUtils.h>
#include "reinforcementRandom.h"
#include <QDebug>
using namespace std;
ReinforcementRandom::ReinforcementRandom()
{
dim = 2;
variance = 0;
bSingleDim = false;
maximum = directions = fvec();
}
ReinforcementRandom::~ReinforcementRandom()
{
}
void ReinforcementRandom::SetParams(float variance, bool bSingleDim)
{
this->variance = variance;
this->bSingleDim = bSingleDim;
}
void ReinforcementRandom::Initialize(ReinforcementProblem *problem)
{
this->problem = problem;
bConverged = false;
dim = problem->gridSize*problem->gridSize;
directions = problem->directions;
maximum = directions;
float value = problem->GetReward(maximum);
maximumValue = 0;
history.push_back(maximum);
HistoryValue().push_back(value);
evaluations = 0;
}
fvec ReinforcementRandom::Update()
{
if(bConverged) return maximum;
int quantizeType = problem->quantizeType;
fvec newSample;
if(variance == 0) // we're doing random search
{
newSample.resize(directions.size());
if(quantizeType) FOR(d, directions.size()) newSample[d] = rand() % (quantizeType==1?9:5);
else FOR(d, directions.size()) newSample[d] = drand48()*2*M_PI;
}
else // we're doing random walk
{
newSample = directions;
fvec randSample; randSample.resize(directions.size());
bool bInsideLimits = true;
if(bSingleDim)
{
// we randomly pick one dimension
int index = rand()%(directions.size());
switch(quantizeType)
{
case 0:
{
newSample[index] += RandN(1, 0, variance)[0];
while(newSample[index] < 0) newSample[index] += 2*M_PI;
while(newSample[index] > 2*M_PI) newSample[index] -= 2*M_PI;
}
break;
case 1:
{
newSample[index] = (int)(newSample[index] + rand()%8) % 9;
}
break;
case 2:
{
newSample[index] = (int)(newSample[index] + rand()%4) % 5;
}
break;
}
}
else
{
switch(quantizeType)
{
case 0:
{
randSample = newSample + RandN(directions.size(), 0, variance);
FOR(d, directions.size())
{
while(randSample[d] < 0) randSample[d] += 2*M_PI;
while(randSample[d] > 2*M_PI) randSample[d] -= 2*M_PI;
bInsideLimits &= randSample[d] >= 0.f && randSample[d] <= 2*M_PI;
}
newSample = randSample;
}
break;
case 1:
{
FOR(d, directions.size()) newSample[d] = (int)(newSample[d] + rand()%8) % 9;
}
break;
case 2:
{
FOR(d, directions.size()) newSample[d] = (int)(newSample[d] + rand()%4) % 5;
}
break;
}
}
}
visited.push_back(newSample);
float value = problem->GetReward(newSample);
evaluations++;
if(value > maximumValue)
{
maximum = newSample;
maximumValue = value;
//qDebug() << "new maximum found at " << maximum[0] << " \tvalue: " << value;
}
directions = maximum;
history.push_back(maximum);
historyValue.push_back(maximumValue);
return newSample;
}
void ReinforcementRandom::Draw(QPainter &painter)
{
int w = painter.viewport().width(), h = painter.viewport().height();
int graphW = 200, graphH = 100, graphPad = 10;
int top = h - 10 - (graphH + 2*graphPad);
int left = 10;
QPainter::RenderHints hints = painter.renderHints();
painter.setRenderHint(QPainter::Antialiasing, false);
QFont font = painter.font();
font.setPointSize(9);
painter.setFont(font);
// we draw the rectangle behind
painter.setOpacity(1);
painter.setBrush(Qt::NoBrush);
painter.setPen(QPen(Qt::black, 2));
painter.drawRect(left, top, graphW + 2*graphPad, graphH + 2*graphPad);
painter.setOpacity(0.6);
painter.setPen(Qt::NoPen);
painter.setBrush(Qt::white);
painter.drawRect(left, top, graphW + 2*graphPad, graphH + 2*graphPad);
painter.setOpacity(1);
painter.setBrush(Qt::black);
painter.setPen(Qt::black);
// we draw the values
double maxValue = -DBL_MAX;
FOR(i, historyValue.size()) maxValue = max(maxValue, historyValue[i]);
int valueLimit = 4;
double upperBound = ((int)ceil(maxValue)/valueLimit + 1)*valueLimit;
painter.setPen(QPen(Qt::black, 2));
QPointF oldPoint;
FOR(i, graphW)
{
int index = i*historyValue.size()/graphW;
QPointF point(i, graphH*(1.f - (historyValue[index]/upperBound)));
point += QPointF(left + graphPad, top + graphPad);
if(i) painter.drawLine(point, oldPoint);
if(i==graphW-1)
{
painter.drawText(point + QPointF(-20,0), QString("%1").arg(historyValue.back(), 0, 'f', 2));
}
oldPoint = point;
}
// we draw the axes
painter.setPen(QPen(Qt::black, 2));
painter.drawLine(left + graphPad, top+graphPad, left + graphPad, top+graphPad + graphH);
painter.drawLine(left + graphPad, top+graphPad+graphH, left + graphPad + graphW, top+graphPad + graphH);
painter.drawText(left + graphPad, top + graphPad, QString("%1").arg(upperBound, 0, 'f', 1));
painter.drawText(left + graphPad, top + graphPad*2 + graphH, QString("0"));
font.setPointSize(9);
painter.setFont(font);
painter.drawText(left, top, graphPad*2 + graphW, graphPad, Qt::AlignCenter, "Maximum Reward");
painter.setRenderHints(hints);
}
const char *ReinforcementRandom::GetInfoString()
{
char *text = new char[1024];
if(variance == 0) sprintf(text, "Random Search");
else sprintf(text, "Random Walk\n");
return text;
}
|
