David Young.
Email:
davidy@sussex.ac.uk;
room: 3R401 (past end of 5C corridor).
The course offers an introduction to key mathematical techniques used in "nouvelle AI".
The course is largely taught in one 2-hour seminar each week. This is on Mondays at 11.30 in room Arundel 1C. For those with relatively little mathematical background, there is an additional 1-2 hour session on Mondays at 2.00 in Mantell 2A4.
Topics in italics are likely to be used to illustrate the mathematical techniques. Not everything will be discussed at the same level of detail. The topics may be changed by agreement with the group.
Course introduction. Self-assessment questionnaire. General discussion of neural network training.
Differential calculus, partial differentiation, summation. Feedforward neural networks and the backpropagation algorithm.
Matrix operations. More network techniques.
Vector geometry. Robot navigation; optic flow.
Numerical methods for integration of differential equations. Robot motion.
Probability. Boltzmann machines; genetic algorithms.
Experimental design, data analysis, hypothesis testing. A-life population studies.
Chaos, fractals and dynamic systems theory.
Fourier transforms and orthogonal basis functions. Signal processing.
The course is assessed by coursework only. For details see the separate document.
The full course notes are available in HTML and in PDF. Further reading is suggested in the appropriate sections of the notes.