Newer
Older
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
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
package cz.vsb.mro0010.neuralnetworks;
import java.util.ArrayList;
import java.util.Arrays;
public class BPNet extends MultiLayeredNet {
protected float error;
public float getError() {
return error;
}
public void setError(float error) {
this.error = error;
}
protected float tolerance;
protected String neuronType;
protected float learnCoeff;
public BPNet( float tolerance, int nrOfLayers, int nrOfInputs, ArrayList<Integer> nrOfNeuronsPerLayer, float slope, float learnCoeff ) {
super(nrOfInputs, nrOfLayers, nrOfNeuronsPerLayer);
this.neuronType = "SigmoidalNeuron";
this.tolerance = tolerance;
this.learnCoeff = learnCoeff;
for (int i = 0; i < nrOfLayers; i++) {
for (int j = 0; j < nrOfNeuronsPerLayer.get(i); j++) {
this.neuronLayers.get(i).add(new SigmoidalNeuron(slope));
}
}
for (int i = 0; i < nrOfLayers; i++) {
this.interconnectionsLayers.add(new InterconnectionsBP(this.learnCoeff));
}
for (Neuron neuronIn : this.inputNeuronLayer) {
for (Neuron neuronFirstLevel : this.neuronLayers.get(0)) {
this.interconnectionsLayers.get(0).addConnection(new Connection(neuronIn, neuronFirstLevel, (float) (Math.random())));
}
}
for (int i = 1; i < nrOfLayers; i++) {
for (Neuron neuronIn : this.neuronLayers.get(i-1)) {
for (Neuron neuronOut : this.neuronLayers.get(i)) {
this.interconnectionsLayers.get(i).addConnection(new Connection(neuronIn, neuronOut, (float) (Math.random())));
}
}
}
}
public float getTolerance() {
return tolerance;
}
public void setTolerance(float tolerance) {
this.tolerance = tolerance;
}
@Override
public String getNeuronType() {
return this.neuronType;
}
@Override
public int learn(String trainingSet) {
boolean learned = false;
int iter = 0;
ArrayList<String> trainingElements = new ArrayList<String>(Arrays.asList(trainingSet.split("\n")));
while(!learned) {
learned = true;
this.error = 0;
for (int i = 0; i < trainingElements.size(); i++) {
learned &= learnStep(trainingElements.get(i));
}
iter++;
// System.out.println(iter);
}
return iter;
// System.out.println("Learned in " + iter + " whole training set iterations.");
}
public boolean learnStep(String trainingElement) {
// Run training Element
String[] splitedTrainingElement = trainingElement.split(" ");
StringBuffer inputString = new StringBuffer();
for (int i = 0; i < this.nrOfInputs; i++) { //Input values
inputString.append(splitedTrainingElement[i]);
inputString.append(" ");
}
ArrayList<Float> expectedValues = new ArrayList<Float>();
for (int i = this.nrOfInputs; i < splitedTrainingElement.length; i++) { //Expected values
expectedValues.add(Float.parseFloat(splitedTrainingElement[i]));
}
this.run(inputString.substring(0, inputString.length() - 1));
// Calculate error
float error = 0;
for (int i = 0; i < expectedValues.size(); i++) {
float y = this.neuronLayers.get(this.nrOfLayers-1).get(i).getState(); //output of ith neuron
float o = expectedValues.get(i);
error += (float)( 0.5 * Math.pow((y-o), 2));
}
if (this.error < error) {
this.error = error;
}
if (error > this.tolerance) { //Error is too high -> modify weights
// Calculate deltas
for (int i = this.nrOfLayers - 1; i >= 0; i -= 1) {
for (Neuron n : this.neuronLayers.get(i)) {
SigmoidalNeuron neuron = (SigmoidalNeuron)n;
if (i == this.nrOfLayers - 1) { //Top layer
float y = neuron.getState();
float o = expectedValues.get(this.neuronLayers.get(i).indexOf(neuron));
float delta = y - o;
neuron.setError(delta);
} else { //Other layers
ArrayList<Connection> connectionsToUpperLayerFromNeuron = new ArrayList<Connection>();
// Find all connections, that have "neuron" as input
for (Connection c : this.interconnectionsLayers.get(i+1).getConnections()) {
if (c.getInputNeuron().equals(neuron))
connectionsToUpperLayerFromNeuron.add(c);
}
float delta = 0;
for (Connection c : connectionsToUpperLayerFromNeuron) {
float deltaUpper = ((SigmoidalNeuron)c.getOutputNeuron()).getError();
float lambdaUpper = ((SigmoidalNeuron)c.getOutputNeuron()).getSlope();
float yUpper = c.getOutputNeuron().getState();
float w = c.getWeight();
delta += deltaUpper*lambdaUpper*yUpper*(1-yUpper)*w;
}
neuron.setError(delta);
}
}
}
// Adjust weights
for (Interconnections interconnectionsLayer : this.interconnectionsLayers) {
interconnectionsLayer.adjustWeights();
}
return false;
} else {
return true;
}
}
public String getOutput() {
StringBuffer output = new StringBuffer();
ArrayList<Neuron> outputLayer = this.neuronLayers.get(this.nrOfLayers-1);
for (int i = 0; i < outputLayer.size(); i++) {
output.append(String.valueOf(outputLayer.get(i).getState()));
output.append(" ");
}
return output.toString();
}
public void changeSlopeTo(float slope) {
for (ArrayList<Neuron> neuronLayer : this.neuronLayers) {
for (Neuron neuron : neuronLayer) {
((SigmoidalNeuron)neuron).setSlope(slope);
}
}
}
public void changeLearnCoeffTo(float learnCoeff) {
for (Interconnections layer : interconnectionsLayers) {
((InterconnectionsBP)layer).setLearningRate(learnCoeff);
}
}
public void resetWeights() {
for (Interconnections layer : interconnectionsLayers) {
for (Connection connection : layer.getConnections()) {
connection.setWeight((float)Math.random());
}
}
}
public void addNeuron(int layerIndex, float slope) {
SigmoidalNeuron newNeuron = new SigmoidalNeuron(slope);
neuronLayers.get(layerIndex).add(newNeuron);
if ((layerIndex < nrOfLayers) && (layerIndex >= 0)) {
Interconnections inputConnectionLayer = this.interconnectionsLayers.get(layerIndex);
if (layerIndex == 0) {
ArrayList<InputLayerPseudoNeuron> inputNeurons = this.inputNeuronLayer;
for (Neuron inputNeuron : inputNeurons) {
inputConnectionLayer.addConnection(new Connection(inputNeuron, newNeuron, (float)Math.random()));
}
} else {
ArrayList<Neuron> inputNeurons = this.neuronLayers.get(layerIndex - 1);
for (Neuron inputNeuron : inputNeurons) {
inputConnectionLayer.addConnection(new Connection(inputNeuron, newNeuron, (float)Math.random()));
}
}
if (layerIndex < nrOfLayers - 1) {
Interconnections outputConnectionLayer = this.interconnectionsLayers.get(layerIndex + 1);
ArrayList<Neuron> outputNeurons = this.neuronLayers.get(layerIndex + 1);
for (Neuron outputNeuron : outputNeurons) {
outputConnectionLayer.addConnection(new Connection(newNeuron, outputNeuron, (float)Math.random()));
}
}
this.nrOfNeuronsPerLayer.set(layerIndex, this.nrOfNeuronsPerLayer.get(layerIndex) + 1 );
} else {
throw new InvalidLayerNumberException();
}
}
public void removeNeuron(int layerIndex) {
int nrOfNeuronsInThisLayer = this.nrOfNeuronsPerLayer.get(layerIndex);
if ((layerIndex < nrOfLayers) && (layerIndex >= 0)) {
if (nrOfNeuronsInThisLayer == 1) {
removeNeuronLayer(layerIndex);
} else {
Neuron removedNeuron = this.neuronLayers.get(layerIndex).get(nrOfNeuronsInThisLayer - 1);
Interconnections inputConnectionLayer = this.interconnectionsLayers.get(layerIndex);
ArrayList<Connection> removedConnections = new ArrayList<Connection>();
for (Connection connection : inputConnectionLayer.getConnections()) {
if (connection.getOutputNeuron().equals(removedNeuron)) {
removedConnections.add(connection);
}
}
for (Connection connection : removedConnections) {
inputConnectionLayer.getConnections().remove(connection);
}
removedConnections = new ArrayList<Connection>();
if (layerIndex < nrOfLayers - 1) {
Interconnections outputConnectionLayer = this.interconnectionsLayers.get(layerIndex + 1);
for (Connection connection : outputConnectionLayer.getConnections()) {
if (connection.getInputNeuron().equals(removedNeuron)) {
removedConnections.add(connection);
}
}
for (Connection connection : removedConnections) {
outputConnectionLayer.getConnections().remove(connection);
}
}
this.neuronLayers.get(layerIndex).remove(removedNeuron);
this.nrOfNeuronsPerLayer.set(layerIndex, this.nrOfNeuronsPerLayer.get(layerIndex) - 1 );
}
} else {
throw new InvalidLayerNumberException();
}
}
public void addNeuronLayer(int nrOfNeurons, int layerIndex, float slope) {
if ((layerIndex < nrOfLayers + 1) && (layerIndex >= 0) && (nrOfNeurons > 0)) {
this.nrOfLayers++;
this.nrOfNeuronsPerLayer.add(layerIndex, nrOfNeurons);
// new layer creation
ArrayList<Neuron> newNeuronLayer = new ArrayList<Neuron>();
for (int i = 0; i < nrOfNeurons; i++) {
newNeuronLayer.add(new SigmoidalNeuron(slope));
}
// old connections removal
if (layerIndex < nrOfLayers - 1) { // only if inner layer is added
this.interconnectionsLayers.remove(layerIndex);
}
// new layer adding
this.neuronLayers.add(layerIndex, newNeuronLayer);
// new connections creation
// input
Interconnections inputConnLayer = new InterconnectionsBP(learnCoeff);
if (layerIndex == 0) {
ArrayList<InputLayerPseudoNeuron> inputNeurons = this.inputNeuronLayer;
ArrayList<Neuron> outputNeurons = newNeuronLayer; //Layers already shifted
for (Neuron inputNeuron : inputNeurons) {
for (Neuron outputNeuron : outputNeurons) {
inputConnLayer.addConnection(new Connection(inputNeuron, outputNeuron, (float)Math.random()));
}
}
} else {
ArrayList<Neuron> inputNeurons = this.neuronLayers.get(layerIndex - 1);
ArrayList<Neuron> outputNeurons = newNeuronLayer; //Layers already shifted, this is new layer
for (Neuron inputNeuron : inputNeurons) {
for (Neuron outputNeuron : outputNeurons) {
inputConnLayer.addConnection(new Connection(inputNeuron, outputNeuron, (float)Math.random()));
}
}
}
this.interconnectionsLayers.add(layerIndex, inputConnLayer);
// output
Interconnections outputConnLayer = new InterconnectionsBP(learnCoeff);
if (layerIndex < nrOfLayers - 1) {
ArrayList<Neuron> inputNeurons = newNeuronLayer;
ArrayList<Neuron> outputNeurons = this.neuronLayers.get(layerIndex + 1); //Layers already shifted
for (Neuron inputNeuron : inputNeurons) {
for (Neuron outputNeuron : outputNeurons) {
outputConnLayer.addConnection(new Connection(inputNeuron, outputNeuron, (float)Math.random()));
}
}
this.interconnectionsLayers.add(layerIndex + 1, outputConnLayer);
}
} else {
throw new InvalidLayerNumberException();
}
}
public void removeNeuronLayer(int layerIndex) {
if ((layerIndex < nrOfLayers ) && (layerIndex >= 0) && (nrOfLayers > 1)) {
// delete output connections
if (layerIndex < nrOfLayers - 1) {
this.interconnectionsLayers.remove(layerIndex + 1);
}
// delete input connections
this.interconnectionsLayers.remove(layerIndex);
// delete neurons on layer
this.neuronLayers.remove(layerIndex);
this.nrOfNeuronsPerLayer.remove(layerIndex);
this.nrOfLayers--;
// create new connections
if (layerIndex < nrOfLayers + 1) {
Interconnections connLayer = new InterconnectionsBP(learnCoeff);
if (layerIndex == 0) {
ArrayList<InputLayerPseudoNeuron> inputNeurons = this.inputNeuronLayer;
ArrayList<Neuron> outputNeurons = this.neuronLayers.get(0);
for (Neuron inputNeuron : inputNeurons) {
for (Neuron outputNeuron : outputNeurons) {
connLayer.addConnection(new Connection(inputNeuron, outputNeuron, (float)Math.random()));
}
}
} else {
ArrayList<Neuron> inputNeurons = this.neuronLayers.get(layerIndex - 1);
ArrayList<Neuron> outputNeurons = this.neuronLayers.get(layerIndex);
for (Neuron inputNeuron : inputNeurons) {
for (Neuron outputNeuron : outputNeurons) {
connLayer.addConnection(new Connection(inputNeuron, outputNeuron, (float)Math.random()));
}
}
}
this.interconnectionsLayers.add(layerIndex, connLayer);
}
} else {
throw new InvalidLayerNumberException();
}
}
@Override
public String toString() {
return getNeuronMap();
}
public String getNeuronMap() {
StringBuffer map = new StringBuffer();
for (int i = 0; i < nrOfLayers; i++) {
map.append(String.valueOf(nrOfNeuronsPerLayer.get(i)));
map.append(" ");
}
map.deleteCharAt(map.length() - 1);
return map.toString();
}
public static void main(String[] args) {
ArrayList<Integer> nrOfNeuronsPerLayer = new ArrayList<Integer>();
nrOfNeuronsPerLayer.add(10);
nrOfNeuronsPerLayer.add(7);
nrOfNeuronsPerLayer.add(2);
BPNet net = new BPNet( (float)0.01, 3, 5, nrOfNeuronsPerLayer, (float)1.8, (float)0.7); // bigger slope = better resolution
String trainingSet = "0.4 0.5 1 0.5 1 0 1\n0 0 0 0 0 1 1\n0.1 0.2 0.3 0.4 0.5 0 0\n1 0 1 0 1 1 0\n0.2 0.4 0 0 0.9 0 1";
net.learn(trainingSet);
net.run("0.4 0.5 1 0.5 1"); //expected 0 1
System.out.println(net.getOutput());
net.run("0 0 0 0 0"); // 1 1
System.out.println(net.getOutput());
net.run("0.1 0.2 0.3 0.4 0.5"); // 0 0
System.out.println(net.getOutput());
net.run("1 0 1 0 1"); // 1 0
System.out.println(net.getOutput());
net.run("0.2 0.4 0 0 0.9"); // 0 1
System.out.println(net.getOutput());
System.out.println("Not trained elements:");
net.run("0.9 0.1 0.9 0.1 0.9"); // expected 1 0
System.out.println(net.getOutput());
net.run("0.01 0.01 0.01 0.01 0.01"); // expected 1 1
System.out.println(net.getOutput());
net.run("0.15 0.15 0.35 0.35 0.5"); // 0 0
System.out.println(net.getOutput());
System.out.println(net.getNeuronMap());
net.addNeuron(0, 1.8f);
System.out.println(net.getNeuronMap());
net.addNeuron(1, 1.8f);
System.out.println(net.getNeuronMap());
net.addNeuron(2, 1.8f);
System.out.println(net.getNeuronMap());
net.removeNeuron(0);
System.out.println(net.getNeuronMap());
net.removeNeuron(1);
System.out.println(net.getNeuronMap());
net.removeNeuron(2);
System.out.println(net.getNeuronMap());
net.addNeuronLayer(5, 0, 1.8f);
System.out.println(net.getNeuronMap());
net.addNeuronLayer(5, 2, 1.8f);
System.out.println(net.getNeuronMap());
net.addNeuronLayer(5, 5, 1.8f);
System.out.println(net.getNeuronMap());
net.removeNeuronLayer(5);
System.out.println(net.getNeuronMap());
net.removeNeuronLayer(2);
System.out.println(net.getNeuronMap());
net.removeNeuronLayer(0);
System.out.println(net.getNeuronMap());
net.learn(trainingSet);
net.run("0.4 0.5 1 0.5 1"); //expected 0 1
System.out.println(net.getOutput());
net.run("0 0 0 0 0"); // 1 1
System.out.println(net.getOutput());
net.run("0.1 0.2 0.3 0.4 0.5"); // 0 0
System.out.println(net.getOutput());
net.run("1 0 1 0 1"); // 1 0
System.out.println(net.getOutput());
net.run("0.2 0.4 0 0 0.9"); // 0 1
System.out.println(net.getOutput());
System.out.println("Not trained elements:");
net.run("0.9 0.1 0.9 0.1 0.9"); // expected 1 0
System.out.println(net.getOutput());
net.run("0.01 0.01 0.01 0.01 0.01"); // expected 1 1
System.out.println(net.getOutput());
net.run("0.15 0.15 0.35 0.35 0.5"); // 0 0
System.out.println(net.getOutput());
}
}