Fyfe C. Artificial Neural Networks and Information Theory

University of Paisley, 2000, -204 pp.This course comprises an advanced course to those new information processing simulations which are intended to emulate the information processors which we find in biology.
Traditional artificial intelligence is based on high-level symbol processing i.e. logic programming, expert systems, semantic nets etc. all rely on there being in existence some high level representation of knowledge which can be manipulated by using some type of formal syntax manipulation scheme - the rules of a grammar. Such approaches have proved to be very successful in emulating human prowess in a number of fields e.g. we now have software which can play chess at Grand Master level; we can match professional expertise in medicine or the law using expert systems; we have software which can create mathematical proofs for solving complex mathematical problems.
Yet there are still areas of human expertise which we are unable to mimic using software e.g. our machines have difficulty reliably reading human handwriting, recognising human faces or exhibiting common sense. Notice how low-level the last list seems compared to the list of achievements: it has been said that the difficult things have proved easy to program whereas the easy things have proved difficult.Information Theory and Statistics
Hebbian Learning
Anti-Hebbian Learning
Objective Function Methods
Identifying Independent Sources
Independent Component Analysis
Learning
Unsupervised Learning using Kernel Methods
A Linear Algebra
B Calculus
C Backpropagation Derivation