Advanced Neural Computing

This course provides both a theoretical and pragmatic overview of neural computing with insights into what makes particular paradigms work in practice.

Syllabus

Day 1 Neural Network Building Blocks and Multi Layer Perceptrons

- Architecture and Notation
- Transfer Functions
- Processing Units
- Softmax Output Function
- Statistical Concepts
- Probability Density Function
- Bayes Theorem
- Maximum Likelihood
- Objective Functions
- Training the Network
- Sensitivity Analysis
- Generalization and Overfitting
- Adaptive Mixtures of Local Experts
- Modular Neural Networks
Lab: Exploring Multi layer Perceptrons
Generalization and Overfitting


Day 2 Time Series Data and Radial Basis Function Networks

- Modeling Time Series Data
- Stationarity
- Trends and Detrending
- Feed Forward Networks
- Werbos Recurrent Network
- Non Stationary Time Series Modeling
- Kalman Training
- Generalized Regression Neural Network
Lab: Time Series Modeling with S&P500 Data
- Radial Basis Functions
- Moody-Darken RBFN
- Architecture
- 3 Phases of Training
- Probabilistic Neural Network
- General Regression Neural Networks
Lab: Radial Basis Function Networks with Leonard-Kramer Fault Diagnosis
Data
General Regression Neural Networks


Day 3 Competitive Networks

- Preliminaries
- Conscience in Competitive Learning
- Counter Propagation Networks
- Learning Vector Quantization
- LVQ1 and LVQ2
- LVQ to PNN
- Self-Organizing Maps
- Architecture
- Training and Neighborhoods
- Prediction and Classification
- Using Output Information in the Input Vector
Lab: LVQ to PNN
Self-Organizing Map with Iris Data


Day 4 Introduction to Genetic Algorithms

- Basic Concepts
- Key Elements
- Encoding and Evaluating a Solution
- Creating New Solutions
- Cross-over
- Mutation
- Incorporating New Solutions into the Data-Base
- Starting the Process
- Summary