Contents -- Acknowledgments -- Preface to the First Edition -- Preface to the Second Edition -- Chapter 1. Introduction and Role of Arti cial Neural Networks -- Chapter 2. Fundamentals of Biological Neural Networks -- Chapter 3. Basic Principles of ANNs and Their Early Structures -- 3.1. Basic Principles of ANN Design -- 3.2. Basic Network Structures -- 3.3. The Perceptron's Input-Output Principles -- 3.4. The Adaline (ALC) -- 3.4.1. LMS training of ALC -- 3.4.2. Steepest descent training of ALC -- Chapter 4. The Perceptron -- 4.1. The Basic Structure -- 4.1.1. Perceptron's activation functions -- 4.2. The Single-Layer Representation Problem -- 4.3. The Limitations of the Single-Layer Perceptron -- 4.4. Many-Layer Perceptrons -- 4.A. Perceptron Case Study: Identifying Autoregressive Parameters of a Signal (AR Time Series Identi cation) -- Chapter 5. The Madaline -- 5.1. Madaline Training -- 5.A. Madaline Case Study: Character Recognition -- 5.A.1. Problem statement -- 5.A.2. Design of network -- 5.A.3. Training of the network -- 5.A.4. Results -- 5.A.5. Conclusions and observations -- 5.A.6. MATLAB code for implementing MADALINE network: -- Chapter 6. Back Propagation -- 6.1. The Back Propagation Learning Procedure -- 6.2. Derivation of the BP Algorithm -- 6.3. Modified BP Algorithms -- 6.3.1. Introduction of bias into NN -- 6.3.2. Incorporating momentum or smoothing to weight adjustment -- 6.3.3. Other modi cation concerning convergence -- 6.A. Back Propagation Case Study: Character Recognition -- 6.A.1. Introduction -- 6.A.2. Network design -- 6.A.3. Results -- 6.A.4. Discussion and conclusions -- 6.A.5. Program Code (C++) -- 6.B. Back Propagation Case Study: The Exclusive-OR (XOR) Problem (2-Layer BP) -- 6.C. Back Propagation Case Study: The XOR Problem

7.2. Binary Hopfield Networks -- 7.3. Setting of Weights in Hop eld Nets - Bidirectional Associative Memory (BAM) Principle -- 7.4. Walsh Functions -- 7.5. Network Stability -- 7.6. Summary of the Procedure for Implementing the Hopfield Network -- 7.7. Continuous Hopfield Models -- 7.8. The Continuous Energy (Lyapunov) Function -- 7.A. Hopfield Network Case Study: Character Recognition -- 7.A.1. Introduction -- 7.A.2. Network design -- 7.A.3. Setting of weights -- 7.A.4. Testing -- 7.A.5. Results and conclusions -- 7.A.6. MATALAB codes -- 7.B. Hopfield Network Case Study: Traveling Salesman Problem -- 7.B.1. Introduction -- 7.B.2. Hopfield neural network design -- 7.B.3. Input selection -- 7.B.4. Implementation details -- 7.B.5. Output results -- 7.B.6. Concluding discussion -- Chapter 8. Counter Propagation -- 8.1. Introduction -- 8.2. Kohonen Self-Organizing Map (SOM) Layer -- 8.3. Grossberg Layer -- 8.4. Training of the Kohonen Layer -- 8.4.1. Preprocessing of Kohonen layer's inputs -- 8.4.2. Initializing the weights of the Kohonen layer -- 8.4.3. Interpolative mode layer -- 8.5. Training of Grossberg Layers -- 8.6. The Combined Counter Propagation Network -- 8.A. Counter Propagation Network Case Study: Character Recognition -- 8.A.1. Introduction -- 8.A.2. Network structure -- 8.A.3. Network training -- 8.A.4. Test mode -- 8.A.5. Results and conclusions -- 8.A.6. Source codes (MATLAB) -- Chapter 9. Adaptive Resonance Theory -- 9.1. Motivation -- 9.2. The ART Network Structure -- 9.3. Setting-Up of the ART Network -- 9.4. Network Operation -- 9.5. Properties of ART -- 9.6. Discussion and General Comments on ART-I and ART-II -- 9.A. ART-I Network Case Study: Character Recognition -- 9.A.1. Introduction -- 9.A.2. The data set -- 9.A.3. Network design -- 9.A.4. Performance results and conclusions -- 9.A.5. Code for ART neural network (Java).

9.B. ART-I Case Study: Speech Recognition -- 9.B.1. Input matrix set-up for spoken words -- 9.B.2. Simulation programs Set-Up -- 9.B.3. Computer simulation of ART program (C-language) -- 9.B.4. Simulation results -- Chapter 10. The Cognitron and the Neocognitron -- 10.1. Background of the Cognitron -- 10.2. The Basic Principles of the Cognitron -- 10.3. Network Operation -- 10.4. Cognitron's Network Training -- 10.5. The Neocognitron -- Chapter 11. Statistical Training -- 11.1. Fundamental Philosophy -- 11.2. Annealing Methods -- 11.3. Simulated Annealing by Boltzman Training of Weights -- 11.4. Stochastic Determination of Magnitude of Weight Change -- 11.5. Temperature-Equivalent Setting -- 11.6. Cauchy Training of Neural Network -- 11.A. Statistical Training Case Study

13.2.1. Basic structural elements -- 13.2.2. Setting of storage weights and determination of winning neurons -- 13.2.3. Adjustment of resolution in SOM modules -- 13.2.4. Links between SOM modules and from SOM modules to output modules -- 13.2.5. Determination of winning decision via link weights -- 13.2.6. Nj weights (not implemented in most applications) -- 13.2.7. Initialization and local minima -- 13.3. Forgetting Feature -- 13.4. Training vs. Operational Runs -- 13.4.1. INPUT WORD for training and for information retrieval -- 13.5. Advanced Data Analysis Capabilities -- 13.5.1. Feature extraction and reduction in the LAMSTAR NN -- 13.6. Correlation, Interpolation, Extrapolation and Innovation-Detection -- 13.6.1. Correlation feature -- 13.6.2. Innovation detection in the LAMSTAR NN -- 13.7. Concluding Comments and Discussion of Applicability -- 13.A. LAMSTAR Network Case Study : Character Recognition -- 13.A.1. Introduction -- 13.A.2. Design of the network -- 13.A.3. Fundamental principles -- 13.A.4. Training algorithm -- 13.A.5. Testing procedure -- 13.A.6. Results and their analysis -- 13.A.7. Summary and concluding observations -- 13.A.8. LAMSTAR CODE (MATLAB) -- 13.B. Application to Medical Diagnosis Problems -- Problems -- References -- Author Index -- Subject Index.