Connectionist Models of Cognition and Perception II

Contents......Page 10
Preface......Page 6
Memory......Page 14
2. Mathematical and Neurocomputational Buffer Models......Page 16
Random-buffer model (RB)......Page 17
Knock-out buffer model (KO)......Page 18
2.2. Neurocomputational Activation-Buffer Model......Page 19
2.3. The Effect of Presentation Rate on Buffer Dynamics......Page 21
2.4. Extending the Knock-Out Buffer......Page 22
3. Selective Updating of the Buffer......Page 24
4. Conclusion......Page 25
References......Page 26
1. Introduction......Page 28
2. Simulation of the Behavioural Experiment......Page 29
2.1. Simulations with a Standard Three-layer Backpropagation Hetero-Associator......Page 30
2.2. Simulations with the Dual Reverberant Memory Self-Refreshing Neural Network (DRSR)......Page 31
3. Conclusion......Page 34
Appendix......Page 36
References......Page 37
Vision......Page 40
1. Introduction......Page 42
2. Temporal coherence of activity levels......Page 44
4. Results and discussion......Page 45
5. Conclusions......Page 48
References......Page 49
1.2. Two methods to compute TTC......Page 52
1.3. Time-to-collision computed through optical parameters......Page 53
1.4. The problem of adaptation......Page 54
1.5. Architecture of the model......Page 55
2.2.3. Generalization test......Page 56
2.3. Discussion......Page 57
3.2. Results......Page 58
4. Conclusions......Page 59
Acknowledgments......Page 60
References......Page 61
Action and Navigation......Page 62
1.1. Sequence Learning......Page 64
1.2. The SRN model of sequence learning......Page 65
2. Forward Models......Page 66
4. A forward model of sequence learning......Page 68
4.1. Manipulating the RSl......Page 70
6. Acknowledgments......Page 72
References......Page 73
1. Introduction......Page 75
2. Simulating Character Production Sequences......Page 76
2.2. Training......Page 77
2.3. Testing......Page 78
2.3.1. Phase One......Page 79
3.1. Phase One......Page 80
3.2. Phase Two......Page 82
4. Summary and Conclusions......Page 83
References......Page 84
An Integration of Two Control Architectures of Action Selection and Navigation Inspired by Neural Circuits in the Vertebrate: The Basal Ganglia Benoit Girard, David Filliat, Jean-Arcady Meyer, Alain Berthoz and Agnes Guillot......Page 85
1. Introduction......Page 86
3. Environment and robot......Page 87
4. Model......Page 88
4.1. Ventral loop......Page 89
5 . Experiments......Page 90
5.1. Topological navigation efficiency......Page 91
5.2. Coordination of the navigation strategies......Page 92
References......Page 93
1. Introduction......Page 95
2.1. Overview of Related Approaches......Page 96
2.2. Learning by Averaging......Page 97
2.3. Staged Learning......Page 98
3.1. Controller Design......Page 99
3.2. Controller Training......Page 100
4. Results......Page 102
5 . Discussion......Page 103
References......Page 104
Developmental Processes......Page 106
1. Introduction......Page 108
2. The Representational Acuity Hypothesis......Page 111
3.1. Cat-Dog Asymmetry......Page 112
3.2. Reversing the Asymmetry......Page 114
3.3. Eliminating the Asymmetry......Page 115
References......Page 116
1. Introduction......Page 118
2. The Basic Set Hypothesis......Page 119
3. Computational Model Based on Reinforcement Learning......Page 120
4.1. Modeling Autism and Williams Syndrome......Page 124
5. Discussion......Page 126
References......Page 127
1. Introduction......Page 128
2. Theoretical perspectives......Page 129
3.2. Gustafsson (1996)......Page 130
4.1. LEABRA......Page 132
4.2. Using LEABRA to model over-specific learning in autism......Page 133
5. Ways forward......Page 134
References......Page 136
Category Acquisition......Page 138
1. Introduction......Page 140
2. Experimental Methods......Page 141
3.1. Entry Level Shift......Page 143
3.2. Network Plasticity......Page 144
3.3. Hidden Unit Activation......Page 145
3.4. Relationship of Variability to Learning......Page 146
4. Conclusions......Page 147
References......Page 148
1. Introduction......Page 150
2.1. Feature Structure of the Model Problem......Page 151
2.2. Empirical Implementation of the Model Problem......Page 153
2.3. A Neural Network Realization......Page 154
3.1. A Neural Network implementation of an Input Activation Search......Page 155
3.2. Experiment I: Input Activation by Color Recall......Page 156
4. Testing Feature Creation by Additional Learning Facilitation......Page 157
5. Discussion......Page 158
References......Page 159
1. Introduction......Page 161
3. Visual data compression with neurocomputational model......Page 162
4. The perceptual model......Page 164
5.1. Back-propagation auto-encoder......Page 165
5.2. Back-propagation heteroassociator......Page 166
6. Conclusion......Page 167
References......Page 168
Attention......Page 170
2. The Control Approach for Attention......Page 172
3. Sensory-Motor Attention Processing......Page 173
4. The CODAM Model of Consciousness......Page 175
5. Evidence for CODAM: Temporal Flow......Page 176
6. Conclusions......Page 178
References......Page 179
1. Introduction......Page 181
2.1. Overview......Page 182
2.2. Feature extraction......Page 183
2.4. Selection network......Page 184
2.6. Rechecking......Page 186
3.2. Visual Search......Page 187
3.3. Prediction and Experiment......Page 189
References......Page 190
1. Introduction......Page 191
2. Background on the Blink and Theoretical Justification for Model......Page 192
3. The Model......Page 194
4. Results......Page 198
6. Acknowledgements......Page 199
References......Page 200
1. Introduction......Page 201
1.1. Dimensional attention......Page 202
1.2. The inverse base-rate effect......Page 203
2.1. Structures and algorithms......Page 204
3.1. Dimensional selective attention......Page 207
3.2. The inverse base-rate effect......Page 208
4. General discussion......Page 209
References......Page 210
High Level Cognition and Implementation Issues......Page 212
1. Introduction......Page 214
1.1. The earlier model......Page 215
2.1. Justification......Page 217
2.2. A new model......Page 218
2.3. Training and testing......Page 220
3. Conclusions and future directions......Page 221
References......Page 222
1. Introduction......Page 224
2. The Evolutionary Models......Page 225
3. Simulation Details......Page 226
4. Simulation Results......Page 228
5. Discussion and Conclusions......Page 232
References......Page 233
1. Introduction......Page 234
2. Multiple Inferences......Page 235
3. Discounting and Motives......Page 236
4.1. Method......Page 237
4.2. Results......Page 238
5. Simulation with a Recurrent Network......Page 239
5.1. Method......Page 240
5.2. Results......Page 241
6. General Conclusion......Page 242
References......Page 243
Approaches to Efficient Simulation with Spiking Neural Networks Colm G. Connolly, Ioana Marian and Ronan G. Reilly......Page 244
1.1. Optimisations at the neural level......Page 245
1.2. The Passage of Time......Page 246
2.1. Reducing the number of events to be delivered......Page 247
2.2. Strategies for an eficient management of the events-list......Page 248
3. Summary......Page 252
References......Page 253
Language and Speech......Page 254
1.1. How much order information is necessary?......Page 256
1.2. Modelling solutions......Page 259
References......Page 264
1.2. Explanation according to the Dual Mechanism model......Page 266
2.1. Experiment 1......Page 267
2.2. Experiment 2......Page 268
2.3. Experiment 3......Page 272
3. Discussion......Page 274
References......Page 275
1. Introduction......Page 276
2. General Description of the Technique......Page 277
3.2. Network Training......Page 279
4.2. Network Training......Page 280
6. Evaluation of the Representations......Page 281
7. Conclusions......Page 282
References......Page 284
1. Introduction......Page 286
2.1. Connectionist models......Page 287
2.2. Other models......Page 288
3.1. A model based on the Self-Organising Map (SOM)......Page 290
4. N-gram based segmentation using the UB strategy......Page 291
5. Experiments......Page 292
6. Discussion......Page 293
References......Page 295
Cognitive Architectures and Binding......Page 296
1. Introduction......Page 298
2. The Model......Page 299
2.1. Object Selection Module......Page 301
2.4. Novelty Detection Module......Page 302
3. A Simulation Example......Page 303
4. Discussion......Page 304
References......Page 306
1. Introduction......Page 308
2. Experiment......Page 309
3.1. SOA......Page 311
3.3. Adjacent and distant errors......Page 312
3.4. Error effects......Page 313
4. Discussion......Page 314
5. The neural model under development......Page 315
6. Conclusion......Page 316
References......Page 317