Computational Neuroscience of Vision

by Rolls, Edmund T.; Deco, Gustavo

• ISBN13: 9780198524885
• ISBN10: 0198524889
• Format: Paperback
• Copyright: 2002-01-10
• Publisher: Oxford University Press

Summary

This exciting new book presents a highly complex subject of vision, focussing on the visual information processing and computational operations in the visual system that lead to representations of objects in the brain. In addition to visual processing, it also considers how visual imputs reachand are involved in the computations underlying a wide range of behaviour, thus providing a foundation for understanding the operation of a number of different brain systems. This fascinating book will be of value to all those interested in understanding how the brain works, and in understandingvision, attention, memory, emotion, motivation and action.

Author Biography

Professor Edmund Rolls DSc is a major figure in the field of neuroscience. His books are usually controversial, but without exception, highly successful. Gustavo Deco is a lesser known scientist, though one with a reputation for high quality research, and unquestionably a future star in psychology/neuroscience

Table of Contents

Introduction	p. 1
Introduction and overview	p. 1
Neurons	p. 2
Neurons in a network	p. 2
Synaptic modification	p. 4
Long-Term Potentiation and Long-Term Depression	p. 7
Distributed representations	p. 11
Definitions	p. 11
Advantages of different types of coding	p. 12
Neuronal network approaches versus connectionism	p. 13
Introduction to three neuronal network architectures	p. 14
Systems-level analysis of brain function	p. 16
The fine structure of the cerebral neocortex	p. 21
The fine structure and connectivity of the neocortex	p. 21
Excitatory cells and connections	p. 21
Inhibitory cells and connections	p. 23
Quantitative aspects of cortical architecture	p. 25
Functional pathways through the cortical layers	p. 27
The scale of lateral excitatory and inhibitory effects, and the concept of modules	p. 29
Backprojections in the cortex	p. 30
Architecture	p. 30
Learning	p. 31
Recall	p. 33
Semantic priming	p. 34
Attention	p. 34
Autoassociative storage, and constraint satisfaction	p. 34
The primary visual cortex	p. 36
Introduction and overview	p. 36
Retina and lateral geniculate nuclei	p. 37
Striate cortex: Area V1	p. 43
Classification of V1 neurons	p. 43
Organization of the striate cortex	p. 45
Visual streams within the striate cortex	p. 48
Computational processes that give rise to V1 simple cells	p. 49
Linsker's method: Information maximization	p. 50
Olshausen and Field's method: Sparseness maximization	p. 53
The computational role of V1 for form processing	p. 55
Backprojections to the lateral geniculate nucleus	p. 55
Extrastriate visual areas	p. 57
Introduction	p. 57
Visual pathways in extrastriate cortical areas	p. 57
Colour processing	p. 61
Trichromacy theory	p. 61
Colour opponency, and colour contrast: Opponent cells	p. 61
Motion and depth processing	p. 65
The motion pathway	p. 65
Depth perception	p. 67
The parietal cortex	p. 70
Introduction	p. 70
Spatial processing in the parietal cortex	p. 70
Area LIP	p. 71
Area VIP	p. 73
Area MST	p. 74
Area 7a	p. 74
The neuropsychology of the parietal lobe	p. 75
Unilateral neglect	p. 75
Balint's syndrome	p. 77
Gerstmann's syndrome	p. 79
Inferior temporal cortical visual areas	p. 81
Introduction	p. 81
Neuronal responses in different areas	p. 81
The selectivity of one population of neurons for faces	p. 83
Combinations of face features	p. 84
Distributed encoding of object and face identity	p. 84
Distributed representations evident in the firing rate distributions	p. 85
The representation of information in the responses of single neurons to a set of stimuli	p. 90
The representation of information in the responses of a population of inferior temporal visual cortex neurons	p. 94
Advantages for brain processing of the distributed representation of objects and faces	p. 98
Should one neuron be as discriminative as the whole organism, in object encoding systems?	p. 103
Temporal encoding in the spike train of a single neuron	p. 105
Temporal synchronization of the responses of different cortical neurons	p. 108
Conclusions on cortical encoding	p. 111
Invariance in the neuronal representation of stimuli	p. 112
Size and spatial frequency invariance	p. 112
Translation (shift) invariance	p. 113
Reduced translation invariance in natural scenes	p. 113
A view-independent representation of objects and faces	p. 115
Face identification and face expression systems	p. 118
Learning in the inferior temporal cortex	p. 120
Cortical processing speed	p. 122
Conclusions	p. 125
Visual attentional mechanisms	p. 126
Introduction	p. 126
The classical view	p. 126
The spotlight metaphor and feature integration theory	p. 126
Computational models of visual attention	p. 129
Biased competition - single cell studies	p. 132
Neurophysiology of attention	p. 133
The role of competition	p. 135
Evidence of attentional bias	p. 136
Non-spatial attention	p. 136
High-resolution buffer hypothesis	p. 139
Biased competition - fMRI	p. 140
Neuroimaging of attention	p. 140
Attentional effects in the absence of visual stimulation	p. 141
The computational role of top-down feedback connections	p. 142
Neural network models	p. 145
Introduction	p. 145
Pattern association memory	p. 145
Architecture and operation	p. 146
The vector interpretation	p. 149
Properties	p. 150
Prototype extraction, extraction of central tendency, and noise reduction	p. 151
Speed	p. 151
Local learning rule	p. 152
Implications of different types of coding for storage in pattern associators	p. 158
Autoassociation memory	p. 159
Architecture and operation	p. 160
Introduction to the analysis of the operation of autoassociation networks	p. 161
Properties	p. 163
Use of autoassociation networks in the brain	p. 170
Competitive networks, including self-organizing maps	p. 171
Function	p. 171
Architecture and algorithm	p. 171
Properties	p. 173
Utility of competitive networks in information processing by the brain	p. 178
Guidance of competitive learning	p. 180
Topographic map formation	p. 182
Radial Basis Function networks	p. 187
Further details of the algorithms used in competitive networks	p. 188
Continuous attractor networks	p. 192
Introduction	p. 192
The generic model of a continuous attractor network	p. 195
Learning the synaptic strengths between the neurons that implement a continuous attractor network	p. 196
The capacity of a continuous attractor network	p. 198
Continuous attractor models: moving the activity packet of neuronal activity	p. 198
Stabilization of the activity packet within the continuous attractor network when the agent is stationary	p. 202
Continuous attractor networks in two or more dimensions	p. 203
Mixed continuous and discrete attractor networks	p. 203
Network dynamics: the integrate-and-fire approach	p. 204
From discrete to continuous time	p. 204
Continuous dynamics with discontinuities	p. 205
Conductance dynamics for the input current	p. 207
The speed of processing of one-layer attractor networks with integrate-and-fire neurons	p. 209
The speed of processing of a four-layer hierarchical network with integrate-and-fire attractor dynamics in each layer	p. 212
Spike response model	p. 215
Network dynamics: introduction to the mean field approach	p. 216
Mean-field based neurodynamics	p. 218
Population activity	p. 218
A basic computational module based on biased competition	p. 220
Multimodular neurodynamical architectures	p. 221
Interacting attractor networks	p. 224
Error correction networks	p. 228
Architecture and general description	p. 229
Generic algorithm (for a one-layer network taught by error correction)	p. 229
Capability and limitations of single-layer error-correcting networks	p. 230
Properties	p. 234
Error backpropagation multilayer networks	p. 236
Introduction	p. 236
Architecture and algorithm	p. 237
Properties of multilayer networks trained by error backpropagation	p. 238
Biologically plausible networks	p. 239
Reinforcement learning	p. 240
Contrastive Hebbian learning: the Boltzmann machine	p. 241
Models of invariant object recognition	p. 243
Introduction	p. 243
Approaches to invariant object recognition	p. 244
Feature spaces	p. 244
Structural descriptions and syntactic pattern recognition	p. 245
Template matching and the alignment approach	p. 247
Invertible networks that can reconstruct their inputs	p. 248
Feature hierarchies	p. 249
Hypotheses about object recognition mechanisms	p. 253
Computational issues in feature hierarchies	p. 257
The architecture of VisNet	p. 258
Initial experiments with VisNet	p. 266
The optimal parameters for the temporal trace used in the learning rule	p. 274
Different forms of the trace learning rule, and their relation to error correction and temporal difference learning	p. 275
The issue of feature binding, and a solution	p. 284
Operation in a cluttered environment	p. 295
Learning 3D transforms	p. 301
Capacity of the architecture, and incorporation of a trace rule into a recurrent architecture with object attractors	p. 307
Vision in natural scenes--effects of background versus attention	p. 313
Synchronization and syntactic binding	p. 319
Further approaches to invariant object recognition	p. 320
Processes involved in object identification	p. 321
The cortical neurodynamics of visual attention - a model	p. 323
Introduction	p. 323
Physiological constraints	p. 324
The dorsal and ventral paths of the visual cortex	p. 324
The biased competition hypothesis	p. 326
Neuronal receptive fields	p. 327
Architecture of the model	p. 328
Overall architecture of the model	p. 328
Formal description of the model	p. 331
Performance measures	p. 336
Simulations of basic experimental findings	p. 336
Simulations of single-cell experiments	p. 337
Simulations of fMRI experiments	p. 339
Object recognition and spatial search	p. 341
Dynamics of spatial attention and object recognition	p. 343
Dynamics of object attention and visual search	p. 345
Evaluation of the model	p. 348
Spatial attention and object attention	p. 348
Translation-invariant object recognition	p. 350
Contributions and limitations	p. 351
Visual search: Attentional neurodynamics at work	p. 353
Introduction	p. 353
Simple visual search	p. 354
Visual search of hierarchical patterns	p. 358
The spatial resolution hypothesis	p. 358
Neurodynamics of the resolution hypothesis	p. 361
Visual search in the framework of the resolution hypothesis	p. 363
Visual conjunction search	p. 369
The binding problem	p. 369
The time course of conjunction search: experimental evidence	p. 371
Extension of the computational cortical architecture	p. 373
Computational results	p. 376
Conclusion	p. 381
A computational approach to the neuropsychology of visual attention	p. 383
Introduction	p. 383
The neglect syndrome	p. 383
A model of visual spatial neglect	p. 384
Spatial cueing effect on neglect	p. 388
Extinction and visual search	p. 390
Effect on neglect of top-down knowledge about objects	p. 392
Hierarchical patterns - neuropsychology	p. 398
Conjunction search - neuropsychology	p. 400
Simulations and predictions	p. 400
Experimental test of the predictions in human subjects	p. 401
Conclusion	p. 403
Outputs of visual processing	p. 404
Visual outputs to Short Term Memory systems	p. 406
Prefrontal cortex short term memory networks, and their relation to temporal and parietal perceptual networks	p. 406
Computational details of the model of short term memory	p. 409
Computational necessity for a separate, prefrontal cortex, short term memory system	p. 412
Role of prefrontal cortex short term memory systems in visual search and attention	p. 412
Synaptic modification is needed to set up but not to reuse short term memory systems	p. 413
Visual outputs to Long Term Memory systems in the brain	p. 413
Effects of damage to the hippocampus and connected structures on object-place and episodic memory	p. 414
Neurophysiology of the hippocampus and connected areas	p. 415
Hippocampal models	p. 418
The perirhinal cortex, recognition memory, and familiarity	p. 421
Visual stimulus-reward association, emotion, and motivation	p. 424
Emotion	p. 425
Reward is not processed in the temporal cortical visual areas	p. 429
Why the reward and punishment associations of stimuli are not represented early in information processing in the primate brain	p. 430
Amygdala	p. 434
Orbitofrontal cortex	p. 439
Effects of mood on memory and visual processing	p. 447
Output to object selection and action systems	p. 448
Visual search	p. 452
Visual outputs to behavioral response systems	p. 453
Multimodal representations in different brain areas	p. 453
Visuo-spatial scratchpad, and change blindness	p. 454
Conscious visual perception	p. 454
Principles and Conclusions	p. 456
Transform invariance in the inferior temporal visual cortex	p. 456
Representation of information in IT	p. 456
IT information processing is fast	p. 457
Continuous neuronal dynamics allows fast network processing	p. 457
Hierarchical feature analysis	p. 457
Trace learning rule for invariant representations	p. 459
Spatial feature binding by feature combination neurons	p. 460
IT provides a representation for later memory networks	p. 461
Face expression and object motion	p. 462
Attentional mechanisms	p. 462
Visual search	p. 464
Egocentric vs allocentric representations	p. 464
Short term memory as the controller of attention	p. 465
Output to object selection and action systems	p. 466
'What' versus 'where' processing streams	p. 466
Short term memory must be separated from perception	p. 467
Backprojections must be weak	p. 468
Long-term potentiation and short-term memory	p. 469
"Executive control" by the prefrontal cortex	p. 469
Reward processing occurs after object identification	p. 470
Effects of mood on memory and visual processing	p. 471
Visual outputs to Long Term Memory systems	p. 471
Episodic memory and the operation of mixed discrete and continuous attractor networks	p. 472
Visual outputs to behavioural response systems	p. 472
Multimodal representations in different brain areas	p. 472
Visuo-spatial scratchpad and change blindness	p. 472
Invariant object recognition and attention	p. 473
Conscious visual perception	p. 473
Attention - future directions	p. 473
Integrated approaches to understanding vision	p. 475
Apostasis	p. 475
Introduction to linear algebra for neural networks	p. 477
Vectors	p. 477
The inner or dot product of two vectors	p. 477
The length of a vector	p. 478
Normalizing the length of a vector	p. 479
The angle between two vectors: the normalized dot product	p. 479
The outer product of two vectors	p. 480
Linear and non-linear systems	p. 481
Linear combinations of vectors, linear independence, and linear separability	p. 482
Application to understanding simple neural networks	p. 484
Capability and limitations of single-layer networks: linear separability and capacity	p. 484
Non-linear networks: neurons with non-linear activation functions	p. 487
Non-linear networks: neurons with non-linear activations	p. 488
Information theory	p. 490
Basic notions	p. 490
The information conveyed by definite statements	p. 491
Information conveyed by probabilistic statements	p. 491
Information sources, information channels, and information measures	p. 492
The information carried by a neuronal response and its averages	p. 494
The information conveyed by continuous variables	p. 496
The information carried by neuronal responses	p. 498
The limited sampling problem	p. 498
Correction procedures for limited sampling	p. 500
The information from multiple cells: decoding procedures	p. 501
Information in the correlations between the spikes of different cells	p. 504
Information theory results	p. 507
Temporal codes versus rate codes within the spike train of a single neuron	p. 507
The speed of information transfer from single neurons	p. 509
The information from multiple cells: independent information versus redundancy across cells	p. 512
The information from multiple cells: the effects of cross-correlations between cells	p. 514
Conclusions	p. 517
Information theory terms--a short glossary	p. 518
References	p. 520
Index	p. 565
Table of Contents provided by Syndetics. All Rights Reserved.