دانلود کتاب Modelling perception with artificial neural networks

فهرست مطالب :

Half-title......Page 3
Title......Page 5
Copyright......Page 6
Contents......Page 7
Contributors......Page 9
Introduction: Modelling perception with artificial neural networks......Page 13
Part I: General themes......Page 17
1.1 Introduction......Page 19
1.2.1 Artificial neurons......Page 20
1.2.2 Feedforward networks and classification......Page 21
1.2.3 Training feedforward networks......Page 25
1.2.4 Networks and generalisation......Page 28
1.2.5 Knowledge representations in neural networks......Page 30
1.2.6 Learning temporal sequences......Page 32
1.3.1 Description versus governance......Page 33
1.3.2 The problem of biological realism......Page 34
1.3.3 Marr’s hierarchical analysis......Page 35
1.3.4 Mapping from networks to biological neural circuits......Page 38
1.3.5 Networks and neural representations......Page 39
1.3.6 Networks and neural architectures......Page 40
1.4 Concluding remarks......Page 42
References......Page 43
2.1 Introduction......Page 47
2.2.1 Vision......Page 49
2.2.1.1 Modifying a visual neural network......Page 51
2.2.3 Audition......Page 53
2.2.4 A simple auditory model......Page 56
2.2.5 Incorporating costs......Page 60
2.3.1 Touch......Page 61
2.3.2 Neural network models of topographic maps......Page 62
2.4 Common themes, uncommon opportunities......Page 64
References......Page 66
Part II: The use of artificial neural networks to elucidate the nature of perceptual processes in animals......Page 73
3.1 Introduction......Page 75
3.3 Results and discussion......Page 77
References......Page 83
4.1 Introduction......Page 86
4.2 Methods......Page 88
4.3 Results......Page 89
4.4 Discussion......Page 95
4.5.1 Model architecture......Page 97
4.5.2 Task and training......Page 98
4.5.5 ROC analysis......Page 101
References......Page 102
5.1 Introduction......Page 105
5.2 The method used: computational embodied neuroscience......Page 108
5.4 The simulated organism, environment and experiments......Page 110
5.5.1 The amygdala: a CS–US associator......Page 113
5.5.2 The cortex-dorsolateral striatum pathway: an S-R associator......Page 115
5.5.3 The amygdala-nucleus accumbens pathway: a bridge between Pavlovian and instrumental processes......Page 116
5.6 Results......Page 117
5.7 Conclusions and future work......Page 121
References......Page 123
6.1 Introduction......Page 126
6.2 The study model: the two cortical visual systems for the ‘What’ and ‘Where’ task......Page 128
6.3.1 Genetic interference and a new hypothesis on the origin of brain architecture......Page 130
6.3.2 Sequential learning of the ‘What’ and ‘Where’ task in nonmodular neural networks......Page 133
6.3.3.1 Generalisation in the ‘What’ and ‘Where’ tasks......Page 135
6.3.3.2 Generalisation in the ecological task......Page 136
6.4 Conclusion: possible improvements of our study model and new simulations......Page 140
References......Page 142
7.1 Introduction......Page 146
7.2 Neural networks and complex networks......Page 148
7.3.1 Hopfield model on a network......Page 149
7.3.2 Storage capacity......Page 150
7.4.1 Results for the network models......Page 151
7.4.2 Results for C. elegans......Page 154
7.5 Analysis......Page 155
7.6 Discussion......Page 158
References......Page 159
8.1 Introduction......Page 161
8.1.1 Brain tumours......Page 162
8.1.2 Tumour-related epilepsy......Page 163
8.1.3 Cognitive functioning in brain tumour patients......Page 164
8.2 Network theory......Page 165
8.2.1 Small-world networks......Page 166
8.2.3 Synchronisation of networks......Page 168
8.3 Neural networks......Page 169
8.3.2 In vitro and in vivo experimental studies......Page 170
8.3.3.1 Functional magnetic resonance imaging......Page 171
8.3.3.2 Electroencephalography (EEG) and magnetoencephalography (MEG)......Page 172
8.4.1 The lesioned brain......Page 174
8.4.2 Brain tumour patients......Page 175
8.5. Cognition and network topology......Page 179
8.5.1 Compensatory mechanisms......Page 180
8.5.2 Patients with brain tumours......Page 181
8.6.1 Simulated, in vitro and in vivo studies......Page 182
8.6.2 Studies in humans......Page 183
8.7 Conclusions and future prospects......Page 185
References......Page 188
Part III: Artificial neural networks as models of perceptual processing in ecology and evolutionary biology......Page 197
9.1 Introduction......Page 199
9.2 Using artificial neural networks to study character displacement......Page 201
9.3 The model......Page 202
9.4 Simulating the evolution of conspecific recognition......Page 205
9.5.1 Stimuli sets......Page 206
9.5.2 Simulations, testing and analyses of networks’ responses......Page 208
9.6.1 Stimuli sets......Page 210
9.6.2 Simulations, testing and analyses of networks’ responses......Page 211
9.7.1 How does character displacement affect preferences for conspecifics and discrimination against heterospecifics?......Page 213
9.7.2 Can reproductive character displacement initiate speciation?......Page 215
9.8 Discussion......Page 220
9.9 Conclusions......Page 223
References......Page 224
10.1 Introduction......Page 227
10.2.1 Concealment......Page 228
10.2.2 Signalling......Page 230
10.2.3 Prey colouration in models......Page 231
10.3.1 Artificial neural networks......Page 232
10.3.2 Prey colouration evolution......Page 234
10.3.3 Neural networks in studies on prey colouration evolution......Page 235
10.4.1 Prey camouflage and visual background complexity......Page 237
10.4.2 The problems of crypsis and aposematism and the choice of defence strategy......Page 239
10.5 Conclusions......Page 242
References......Page 245
11.1 Introduction......Page 248
11.2 Recognition system and ANN design......Page 249
11.3 Stimulus-response functions and specialisation......Page 251
11.4.1 The ideal free distribution......Page 252
11.4.2 The evolution of guilds......Page 254
11.5 Specialisation when resources evolve......Page 255
11.5.2 Mimicry evolution......Page 256
11.6 Exploiter asexual versus sexual reproduction......Page 258
11.7 Conclusions......Page 261
11.8.1 The exploiters......Page 263
Acknowledgement......Page 264
References......Page 265
12.1 Stochastic multi-modal communication......Page 267
12.2 Receiving as a feedforward pathway......Page 269
12.2.1 Information measures......Page 270
12.3 Network complexity measures and multi-modal modules......Page 271
12.4.1 A definition of robustness......Page 272
12.4.2 An example: duplication and robustness......Page 273
12.4.3 Extension of robustness measure to recurrent networks......Page 274
12.5 Multi-modality and combinatoriality......Page 277
12.7 Signalling and evolutionary innovation......Page 279
References......Page 280
13.1 Introduction......Page 281
13.2.1.2 Plant motion backgrounds......Page 284
13.2.1.4 Image motion analysis......Page 285
13.2.2.1 Feature and conspicuity maps......Page 286
13.2.2.2 The saliency map and winner-take-all neural network......Page 288
13.3 Results and discussion......Page 289
13.3.1 Signal efficacy and increased plant motion noise......Page 291
13.3.2 Signal efficacy and signaller–plant distance......Page 294
13.3.3 Does signalling faster help?......Page 296
13.4 General discussion and outlook......Page 300
References......Page 302
Part IV: Methodological issues in the use of simple feedforward networks......Page 305
14.1 Introduction......Page 307
14.2.2 Learning......Page 308
14.2.3 Model stimuli and generalisation tests......Page 309
14.3.1 Errorless discrimination learning......Page 310
14.3.2 Disappearance of peak shift in extinction......Page 312
14.3.3 Range and frequency effects......Page 313
14.4 Discussion......Page 315
14.5.1 Stimuli......Page 316
14.5.2 Training and testing......Page 317
References......Page 318
15.1 Introduction......Page 320
15.2.1 The neural network and training procedure......Page 321
15.2.3 Characterising network error-weight surfaces......Page 322
15.3 Results and discussion......Page 323
15.3.1 The need for stochastic replication of ecological neural networks......Page 327
References......Page 328
16.1 Introduction......Page 330
16.2 What do we mean by ecological learning?......Page 332
16.3 Why neural networks as they are commonly implemented are not good models of ecological learning......Page 333
16.4.1 Methods......Page 335
16.4.2 Results......Page 336
16.5 A neural network with some properties analogous to ecological learning......Page 337
16.5.1 Methods......Page 338
16.5.2 Results......Page 339
16.6 Discussion......Page 340
References......Page 344
17.2 The evolution of neural networks......Page 346
17.2.2 Evolution of weights......Page 347
17.2.3 Evolution of architectures......Page 348
17.2.7 Comparison of approaches......Page 349
17.3.1.1 Connectionist encoding......Page 350
17.3.1.4 Layer-based encoding......Page 352
17.3.2 Indirect encoding......Page 353
17.3.2.3 Edge encoding......Page 354
17.4 Comparison of encoding schemes......Page 355
17.5.1 Examining the effects of imitation on populations of evolutionary neural networks......Page 356
17.5.2 Insect path integration......Page 358
17.6 Chapter summary......Page 359
References......Page 360
18.1 Introduction......Page 363
18.2.1 Study organisms......Page 365
18.2.3 Movement experiments......Page 367
18.2.5 Multi-response artificial neural network......Page 369
18.2.6 Quantifying variable importance in neural networks......Page 371
18.2.6.2 Connection weight approach......Page 372
18.3 Results......Page 373
18.4 Discussion......Page 378
18.5 Conclusion......Page 381
References......Page 382
19.1 Introduction......Page 386
19.2.1 Gross morphology......Page 387
19.2.2 Crista morphology......Page 388
19.3.1 Peripheral connections and responses......Page 389
19.3.2 The peripheral efferent system......Page 391
19.3.3 Extensive electrical coupling within the crista network......Page 393
19.3.5 The central afferent and efferent projections......Page 394
19.4.1 The primary sensory hair cell network system......Page 395
19.4.2 The secondary sensory hair cell network system......Page 396
19.4.3 System complexity and flexibility......Page 397
19.4.4 Network considerations......Page 398
References......Page 399
Index......Page 402