Natural computation for intell. sys. – Page 2

Category: Natural computation for intell. sys.

This unit looks at the development and application of biologically inspired models of computation. We study: basic components of a natural neural systems: synapses, dendrites and neurons and their computational models; fundamental concepts of data and signal encoding and processing; neural network architectures: pattern association networks, auto associative networks, feedforward networks, competitive networks, self organizing networks and recurrent networks; plasticity and learning. Hebb rule, supervised learning, reinforced learning, error-correcting learning, unsupervised learning, competitive learning, self-organization.

FIT5167 – Natural Computation Week 1

Natural computation for intelligent systems is a subject dedicated to the use of Neural Networks for the use of pattern recognition. Of the four subjects introductory lectures I found Natural Computation to be the most stressful. We were hit with a ‘self-assessment’ math test which we were advised would be the taste of the course. It has been a long time since I have done pure mathematics so was not completely comfortable with some topics covered.

Assuming that a great deal of pure mathematics will be required to do well in this subject it will likely be my most difficult this semester.

I have begun reading the prescribed text (S. Samarasinghe, Neural Networks for Applied Sciences and Engineering) however I still feel very foggy about a lot of the concepts raised and about neural networks in general.

The key advantages of neural networks appears to be the high degree of parallelism and the ability to ‘learn’.

Learning can be accomplished through supervised or unsupervised methods but these seem to always be decided by the application rather than the implementer.

This week I will get familiar with MatLab and hope to find some software that can emulate neural networks.