NEURAL NETWORKS by Hervé Abdi, Dominique Valentin, and Betty Edelman

Abdi, H., Valentin, D., Edelman, B. (1999). Neural Networks.
Thousand Oaks (CA): Sage. (100p.)
Quantitative Applications in the Social Sciences Series # 124.
ISBN 0-7619-1440-4. Price: $13.95.


Contents:

1. Series Editor's Introduction v
2. Notations vi
3. Introduction 1
   1. What are neural networks? 1
   2. Overview of this book 2
4. The Perceptron 3
   1. Overview 3
   2. The McCulloch and Pitts neuron 4
   3. Architecture of a perceptron 8
   4. Widrow-Hoff learning rule 9
   5. Learning with +1/-1 cells 13
   6. Performance evaluation 17
   7. Perceptron and discriminant analysis 19
   8. Learning and testing sets: The validation problem 20
5. Linear Autoassociative Memories 21
   1. Overview 21
   2. The building block: the basic linear unit 22
   3. Architecture of an autoassociative memory 23
   4. Hebbian learning rule 25
   5. Retrieval of a learned pattern 29
   6. Limitations of Hebbian Learning 31
   7. Generalization to new stimuli 35
   8. Types of errors 36
   9. The Widrow-Hoff learning rule 37
   10. Singular value decomposition: PCA models 41
   11. The Widrow-Hoff rule and gradient descent 45
6. Linear Heteroassociative Memories 46
   1. Overview 46
   2. Architecture of a heteroassociative memory 47
   3. Hebbian learning rule 48
   4. Widrow-Hoff learning rule 54
   5. Widrow-Hoff learning and pseudo-inverse 58
   6. Widrow-Hoff rule and gradient descent 60
   7. Discriminant analysis and perceptron revisited 60
   8. Radial basis function networks 61
7. Error Backpropagation 68
   1. Overview 68
   2. Architecture and notation 69
   3. The building block: nonlinear units 70
   4. The backpropagation algorithm 73
   5. Performance analysis 83
   6. Backpropagation and gradient descent 85
   7. Backpropagation and logistic regression 87
8. Useful References 87
   1. References 88
9. About the Authors 90

back to Hervé Abdi's home page

Link to Sage home page for order information