|
|
Foreword--vii
Preface--xi
List of Figures--xxv
List of Tables--xxxi
Part I FOUNDATIONS
1 Introduction -- 3
1.1 Input-Driven Self-Organization -- 4
1.2 Constructing Visual Function -- 7
1.3 Perceptual Grouping -- 8
1.4 Approach -- 10
1.5 Guide for the Reader -- 13
2 Biological Background -- 15
2.1 Visual System Organization -- 15
2.2 Lateral Connections -- 23
2.3 Genetic Versus Environmental Factors in Development -- 28
2.4 Temporal Coding -- 33
2.5 Conclusion -- 36
3 Computational Foundations -- 39
3.1 Computational Units -- 39
3.2 Temporal Coding -- 46
3.3 Adaptation -- 48
3.4 Self-Organizing Maps -- 50
3.5 Knowledge Representation in Maps -- 57
3.6 Conclusion -- 63
Part II INPUT-DRIVEN SELF-ORGANIZATION
4 LISSOM: A Computational Map Model of V1 -- 67
4.1 Motivation: Cortical Maps -- 67
4.2 The LISSOM Architecture -- 68
4.3 Response Generation -- 73
4.4 Learning -- 76
4.5 Self-Organizing Process -- 78
4.6 Conclusion -- 82
5 Development of Maps and Connections -- 85
5.1 Biological Background -- 85
5.2 Computational Models -- 91
5.3 Orientation Maps -- 95
5.4 Ocular Dominance Maps -- 106
5.5 Direction Selectivity Maps -- 113
5.6 Combined Maps of Multiple Features -- 119
5.7 Discussion -- 130
5.8 Conclusion -- 132
6 Understanding Plasticity -- 133
6.1 Biological and Computational Background -- 133
6.2 The Reduced LISSOM Model -- 138
6.3 Retinal Lesions -- 142
6.4 Cortical Lesions -- 146
6.5 Discussion -- 152
6.6 Conclusion -- 153
7 Understanding Visual Performance: The Tilt Aftereffect -- 155
7.1 Psychophysical and Computational Background -- 155
7.2 Method -- 160
7.3 Results -- 164
7.4 Analysis -- 166
7.5 Discussion -- 170
7.6 Conclusion -- 171
Part III CONSTRUCTING VISUAL FUNCTION
8 HLISSOM: A Hierarchical Model -- 175
8.1 Motivation: Synergy of Nature and Nurture -- 175
8.2 The Hierarchical Architecture -- 178
8.3 Inputs, Activation and Learning -- 181
8.4 Effect of Input Sequence and Initial Organization -- 184
8.5 Conclusion -- 185
9 Understanding Low-Level Development: Orientation Maps -- 189
9.1 Biological Motivation -- 189
9.2 Prenatal Development -- 190
9.3 Postnatal Development -- 194
9.4 Prenatal and Postnatal Contributions -- 199
9.5 Discussion -- 201
9.6 Conclusion -- 202
10 Understanding High-Level Development: Face Detection -- 203
10.1 Psychophysical and Computational Background -- 203
10.2 Prenatal Development -- 213
10.3 Postnatal Development -- 227
10.4 Discussion -- 233
10.5 Conclusion -- 238
Part IV PERCEPTUAL GROUPING
11 PGLISSOM: A Perceptual Grouping Model -- 241
11.1 Motivation: Temporal Coding -- 241
11.2 The Self-Organization and Grouping Architecture -- 243
11.3 Spiking Unit Model -- 244
11.4 Learning -- 247
11.5 Self-Organizing Process -- 249
11.6 Conclusion -- 255
12 Temporal Coding -- 257
12.1 Method -- 257
12.2 Binding Through Synchronization -- 258
12.3 Segmentation Through Desynchronization -- 263
12.4 Robustness Against Variation and Noise -- 265
12.5 Discussion -- 270
12.6 Conclusion -- 271
13 Understanding Perceptual Grouping: Contour Integration -- 273
13.1 Psychophysical and Computational Background -- 273
13.2 Contour Integration and Segmentation -- 280
13.3 Contour Completion and Illusory Contours -- 288
13.4 Influence of Input Distribution on Anatomy and Performance -- 295
13.5 Discussion -- 299
13.6 Conclusion -- 304
Part V EVALUATION AND FUTURE DIRECTIONS
14 Computations in Visual Maps -- 307
14.1 Visual Coding in the Cortex -- 307
14.2 Visual Coding in LISSOM -- 309
14.3 Visual Coding for High-Level Tasks -- 314
14.4 Discussion -- 319
14.5 Conclusion -- 324
15 Scaling LISSOM simulations -- 325
15.1 Parameter Scaling Approach -- 325
15.2 Scaling Equations -- 326
15.3 Forming Large Maps: The GLISSOM Approach -- 332
15.4 GLISSOM Scaling -- 332
15.5 Scaling to Cortical Dimensions -- 339
15.6 Discussion -- 342
15.7 Conclusion -- 343
16 Discussion: Biological Assumptions and Predictions -- 345
16.1 Self-Organization -- 345
16.2 Genetically Driven Development -- 354
16.3 Temporal Coding -- 358
16.4 Predictions -- 362
16.5 Conclusion -- 373
17 Future Work: Computational Directions -- 375
17.1 Extensions to the LISSOM Mechanisms -- 375
17.2 Modeling New Phenomena with LISSOM -- 379
17.3 New Research Directions -- 396
17.4 The Topographica Cortical Map Simulator -- 403
17.5 Conclusion -- 406
18 Conclusion -- 409
18.1 Contributions -- 409
18.2 Conclusion -- 412
Appendices
A LISSOM Simulation Specifications -- 415
A.1 Generalized Activation Equation -- 415
A.2 Default Parameters -- 416
A.3 Choosing Parameters for New Simulations -- 419
A.4 Retinotopic Maps -- 422
A.5 Orientation Maps -- 422
A.6 Ocular Dominance Maps -- 423
A.7 Direction Maps -- 424
A.8 Combined Orientation / Ocular Dominance Maps -- 424
A.9 Combined Orientation / Ocular Dominance / Direction Maps -- 424
B Reduced LISSOM Simulation Specifications -- 427
B.1 Plasticity -- 427
B.2 Tilt Aftereffect -- 428
B.3 Scaling -- 428
C HLISSOM Simulation Specifications -- 429
C.1 V1 Only -- 429
C.2 Face-Selective Area Only -- 430
C.3 Combined V1 and Face-Selective Area -- 432
D PGLISSOM Simulation Specifications -- 435
D.1 Self-Organization -- 435
D.2 Grouping -- 437
D.3 Synchronization -- 438
E SOM Simulation Specifications -- 439
F Visual Coding Simulation Specifications -- 441
F.1 Sparse Coding and Reconstruction -- 441
F.2 Handwritten Digit Recognition -- 442
G Calculating Feature Maps -- 445
G.1 Preference Map Algorithms -- 445
G.2 Retinotopic Maps -- 449
G.3 Orientation Maps -- 449
G.4 Ocular Dominance Maps -- 449
G.5 Direction Maps -- 450
G.6 Orientation Gradients -- 450
References--451
Author Index--503
Subject Index--523
|
|
