CPEN 502

2Architecture for Learning Systems

Introduction to learning in neural networks, error backpropagation, simulated annealing, content addressable memories.  Data representation topics.  Reinforcement learning (RL).  Implementation challenges in real world scale problems.  Architectures for function approximation in RL.  Comparison with conventional AI:  history and emerging trends.

3 credits

Course Description

Welcome to the exciting world of Machine Learning! Did you know that the best Backgammon players in the world are computers? Furthermore, did you know that no human taught them to play, that instead they taught themselves!

This course is about machine learning with emphasis on artificial neural networks (ANN). We examine a number of learning paradigms and demonstrate how to implement and where best to apply a number of ANN techniques such as Error-Backpropagation, Hopfield Nets and the Self Organising Feature Map. The course is quite practical in nature and by the end you will have learned how to build your own learning agent! We will also cover Reinforcement Learning (RL) and see how it can be combined with neural networks to build a powerful AI. To provide context, these approaches are compared with more traditional forms of machine learning.


This course is intended for anyone interested in the branch of artificial intelligence known as neural networks. There are no pre-requisites for the course other than a reasonable grasp of engineering mathematics. If you would like to know more about neural networks, what they are and how to apply them, then this course is for
you. It will provide a comprehensive and practical introduction to this exciting field! To complete the course you are required to undertake a practical assignment and submit a project report. The practical work will involve Java programming. Although not a requirement, some proficiency with Java programming will come in handy.


Fausett, L. (1994). Fundamentals of Neural Networks. Architectures, Algorithms and Applications. Prentice Hall.

Sutton R.S., and Barto A.G. (1998) Reinforcement Learning. The MIT Press.

More Information

UBC Course Page