Neural Network design and Machine Learning

3 credit points

Synopsis

This course focuses on theory and the most commonly used methods for neural network design, and their relation to machine learning. After the course, students will learn to design suitable neural networks in pattern recognition and nonlinear system modelling, or improve their design in these applications.

The course is suitable for graduate students, especially for students who is engaged in or will be engaged in the study of neural networks and machine learning field.

Outcomes

Upon completion of the course, students are able to understand basic artificial neural models and their design methods, including

lmaster the principle of Multi-layer Perceptrons and Radial Basis Function Networks, and the basic and improved parameter learning algorithms;

lmaster the non-structure design methods of common feed-forward networks including optimal stopping, active learning, neural network ensembles, etc.;

lmaster the structure design methods of common feed-forward networks including pruning, constructing, and evolution algorithms;

lunderstand what and how those factors affect the generalization ability of a neural network;

lunderstand the principle and design methods of some common feed-back neural networks including Hopfield neural networks and self-organizing feature map;

luse neural networks to solve problems such as pattern recognition and function approximation.

Assessment

There will be two in-term projects, and a final the last day of class.

You are required to complete two class projects. The choice of the topic is up to you so long as it clearly pertains to the course materials. We expect a five page write-up about each of the projects, which should clearly and succinctly describe the project goal, methods, and your results.

Your overall grade will be determined roughly as follows:

Chief examiner(s)

Professor Haikun Wei