Neural Networks for Perception

Opis

Neural Networks for Perception, Volume 1: Human and Machine Perception focuses on models for understanding human perception in terms of distributed computation and examples of PDP models for machine perception.

This book addresses both theoretical and practical issues related to the feasibility of both explaining human perception and implementing machine perception in terms of neural network models. The book is organized into two parts. The first part focuses on human perception. Topics on network model of
object recognition in human vision, the self-organization of functional architecture in the cerebral cortex, and the structure and interpretation of neuronal codes in the visual system are detailed under this part. Part two covers the relevance of neural networks for machine perception. Subjects considered under this section include the multi-dimensional linear lattice for Fourier and Gabor transforms, multiple- scale Gaussian filtering, and edge detection; aspects of invariant pattern and object recognition; and neural network for motion processing.

Neuroscientists, computer scientists, engineers, and researchers in artificial intelligence will find the book useful.

Spis treści

Front Cover 2
Human and Machine Perception 5
Copyright Page 6
Table of Contents 9
Contents of Volume 2 13
Contributors 15
Foreword 19
PART I: Human Perception 25
- I. Introduction 27
  - Chapter 1.1. Visual Cortex: Window on the Biological Basis of Learning and Memory 32
    - I. Introduction 32
    - II. Discussion 45
    - References 47
  - Chapter 1.2. A Network Model of Object Recognition in Human Vision 49
    - Abstract 49

Front Cover 2
Human and Machine Perception 5
Copyright Page 6
Table of Contents 9
Contents of Volume 2 13
Contributors 15
Foreword 19
PART I: Human Perception 25
- I. Introduction 27
  - Chapter 1.1. Visual Cortex: Window on the Biological Basis of Learning and Memory 32
    - I. Introduction 32
    - II. Discussion 45
    - References 47
  - Chapter 1.2. A Network Model of Object Recognition in Human Vision 49
    - Abstract 49
    - 1 Introduction 50
    - 2 Shape-Based Recognition Performance in Human Vision 51
    - 3 The CLF Model 55
    - References 62
    - 4 Discussion 62
  - Chapter 1.3. A Cortically Based Model for Integration in Visual Perception 65
    - 1. Introduction 65
    - II. The RCI Model 69
    - IIΙ. Implications for Perception 83
    - Acknowledgements 84
    - References 85
  - CHAPTER 1.4. THE SYMMETRIC ORGANIZATION OF PARALLEL CORTICAL SYSTEMS FOR FORM AND MOTION PERCEPTION 88
    - 1. Introduction: Why Do Parallel Cortical Systems Exist for the Perception of Static Form and Moving Form? 88
    - II. The Motion BCS and the Static BCS 89
    - III. Joining Sensitivity to Direction-of-Motion with Insensitivity to Direction-of-Contrast 91
    - IV. Why is not a Motion Form Perception System Sufficient? 95
    - V. A Symmetry Principle for Cortical Development: Sustained-Transient Gating, Opponent Processing, and Insensitivity to Directio 95
    - VI. Different Geometries and Afterimages for Perception of Static Form and Motion Form 96
    - VII. Resonance versus Reset: Cooperative Feature Linking without Destructive Smearing 96
    - VIII. Combining Rapid Reset and Spatial Impenetrability Predictions 97
    - IX. Perception of Moving-Form-in-Depth: The V1→V2→MT Pathway 98
    - X. Apparent Motion of Illusory Figures 99
    - XI. Design of a MOC Filter 100
    - XII. Continuous Motion Paths from Spatially Stationary Flashes 106
    - ΧΙII. Feature Integration and Spatial Attention Shifts 111
    - XIV. Augmenting the Static BCS 112
    - XV. Design of Simple On-Cells and Off-Cells in the SOC Filter 114
    - ΧVI. FM Symmetry 115
    - XVII. 90° Orientations: From V1 to V2 116
    - XVIII. 180° Opponent Directions from V1 to MT 118
    - XIX. Opponent Rebounds: Rapid Reset of Resonating Segmentations 118
    - XX. MacKay Afterimages, the Waterfall Effect, and Long Range Motion Aftereffects 119
    - XXI. Concluding Remarks: FACADE Principles of Uncertainty, Complementarity, Summetry, and Resonance as a Foundation for Biologic 121
    - References 123
  - Chapter 1.5. The Structure and Interpretation of Neuronal Codes in the Visual System 128
    - 1. Defining the stimulus-response relation 129
    - 2. Stimulus parameter interactions 133
    - 3. The multiplex-filter hypothesis 135
    - 4. Decoding multiplex messages 137
    - 5. Feature selection 139
    - 6. Feature linking 140
    - 7. Interpreting Neuronal Codes 141
    - 8. Conclusion 142
    - REFERENCES 142
  - Chapter 1.6. Self-Organization of Functional Architecture in The Cerebral Cortex 144
    - I. Introduction 144
    - II. Self-Organization of Synaptic Connections 145
    - III. Receptive Fields 151
    - IV. Self-Organization of Receptive Fields 153
    - V. Simple Application 157
    - VI. Summary 162
    - Appendix 1 163
    - Appendix 2 164
    - Appendix 3 166
    - References 167
  - Chapter 1.7. Filters Versus Textons in Human and Machine Texture Discrimination 169
    - I. Abstract 169
    - II. Introduction 170
    - III. Perceptual Asymmetries 175
    - IV. Conclusions 194
    - References 198
  - Chapter 1.8. Two-Dimensional Maps and Biological Vision: Representing Three-Dimensional Space 200
    - 1. Introduction 200
    - 2. Biological Vision 201
    - 3. Computational Vision 204
    - 4. Conclusion 213
    - References 214
  - PART II: Machine Perception 217
    - II. Introduction 219
      - II.1. WISARD and other Weightless Neurons 226
      - I. Introduction 226
      - II. The RAM Node 227
      - III. Generalizing RAM Nodes 229
      - IV. Large Neurons 230
      - V. Associative Networks: A Memorandum 236
      - References 236
    - Chapter II.2. Multi-Dimensional Linear Lattice for Fourier and Gabor Transforms, Multiple-Scale Gaussian Filtering, and Edge Det 238
      - I. Introduction 238
      - II. Successive Convolution in terms of the Central Limit Theorem 242
      - III. One dimensional Gaussian Convolution 245
      - IV. Two-Dimensional Gaussian Averaging 247
      - V. Three and Higher Dimensional Averaging 249
      - VI. Higher order Webs 249
      - VII. Multi-Scale Edge masks 250
      - VIII. Deriving the Gaussian-Smoothed-Laplacian 251
      - IX. Generating the Discrete Fourier Transform 252
      - X. Gabor Transforms 253
      - XI. Simulation Results 254
      - References 256
      - XII. Discussion 256
    - CHAPTER II.3. ASPECTS OF INVARIANT PATTERN AND OBJECT RECOGNITION 258
      - 1. Introduction 258
      - II. Pattern Recognition 260
      - III. Brief Notes in Object Recognition 269
      - IV. Conclusion 270
      - REFERENCES 271
    - CHAPTER II.4. A NEURAL NETWORK ARCHITECTURE FOR FAST ON-LINE SUPERVISED LEARNING AND PATTERN RECOGNITION 272
      - I. Introduction 272
      - II. The ARTMAP System 274
      - III. ARTMAP Simulations: Distinguishing Edible and Poisonous Mushrooms 275
      - IV. Conclusion: Predictive ART 283
      - Acknowledgements 287
      - References 287
    - Chapter II.5. Neural Network Approaches to Color Vision 289
      - I. Introduction 289
      - II. Learning Lightness Algorithms 290
      - III. Finding Material Boundaries: The Use of Color in Image Segmentation 299
      - References 307
    - Chapter II.6. Adaptive Sensory-Motor Coordination Through Self-Consistency 309
      - 1. Introduction 309
      - II. Locating Stationary Targets 312
      - III. Reaching stationary targets with multiple joints 324
      - IV. Dynamic movement control for unforeseen payloads 329
      - V. Summary 336
      - Appendix 337
      - References 337
    - Chapter II.7. Finding Boundaries in Images 339
      - I. Introduction 339
      - II. Local analysis of image patches by filtering 344
      - III. Texture Discrimination 349
      - IV· Brightness boundaries 356
      - Acknowledgements 361
      - References 364
    - Chapter II.8. Compression of Remotely Sensed Images using Self Organizing Feature Maps 369
      - I. Introduction 369
      - II. Self Organizing Feature Map (SOFM) 373
      - III. MasPar Implementation of Kohonen's SOFM 376
      - IV. Results and Discussion 381
      - References 383
      - Acknowledgments 383
    - Chapter II.9. Self - Organizing Maps and Computer Vision 392
      - 1. Introduction 392
      - 2.The Self-Organizing Feature Map and Feature Detection 396
      - 3. Texture segmentation by the SOFM 400
      - 4. Extracting global curve features by the SOFM 404
      - 5. Conclusions 407
      - References 408
    - Chapter II.10. Region Growing Using Neural Networks 410
      - I. Introduction 410
      - II. Region Growing Using Neural Networks 412
      - III. Experimental Results 413
      - IV. Discussion 418
      - References 420
      - V. Summary 420
    - Chapter II.11. Vision and Space-Variant Sensing 422
      - Abstract 422
      - 1. Introduction 423
      - 2. Fixation and space-variant sensing 425
      - 3. Conclusion 442
      - 4. References 446
    - Chapter II.12. Learning and Recognizing 3D Objects from Multiple Views in a Neural System 450
      - I Introduction 450
      - II Background 451
      - Ill Neural System Architecture 452
      - IV System Neurodynamics 456
      - V Demonstration 463
      - VI Conclusion 466
      - References 467
    - Chapter II.13. Hybrid Symbolic-Neural Methods for Improved Recognition Using High-Level Visual Features 469
      - I. Introduction 469
      - II. The KBANN Algorithm 471
      - III. Empirical Tests of KBANN-DAID 476
      - IV. Discussion 482
      - V. Related Work 483
      - VI. Conclusions 484
      - Acknowledgement 484
      - References 484
    - Chapter II.14. Multiscale and Distributed Visual Representations and Mappings for Invariant Low-Level Perception 486
      - I. Introduction 486
      - II. Multiscale and Distributed Image Representations 489
      - III. Invariance 491
      - IV. Mappings 494
      - V. Conclusions 496
      - References 497
    - Chapter II.15. Symmetry: a context free Cue for Foveated Vision 501
      - 1 Introduction 501
      - 2 Defining tlie Operator 503
      - 3 Operation on Natural scenes 506
      - 4 Psychophysical Correlates 509
      - 5 Implementation by Neural Networks 511
      - 6 Conclusion 513
      - References 514
    - Chapter II.16. A Neural Network for Motion Processing 516
      - Abstract 516
      - 1. Introduction 517
      - 2. A Neural Network 518
      - 3. Computing Optical Flow 522
      - 4. Recovering Depth 532
      - 5. Discussion 537
      - References 539
    - Index 541

Pokaż więcej

Human and Machine Perception

Opis

Spis treści

Recenzje (0)