CIS 525: NEURAL COMPUTATION
Fall 2006
Time: Tuesday,
4:40-7:20pm; Place: Tuttleman 305 A
Instructor: Zoran Obradovic
303 Wachman Hall,
Office hours: Tuesday
Course materials: www.ist.temple.edu/~zoran/teaching/cis525.htm
Goals:
Neural
networks provide powerful techniques to model and control nonlinear and complex
systems. The course is designed to provide an introduction to this
interdisciplinary topic. The course is structured such that students from
computer science, engineering, physics, mathematics, statistics, cognitive
sciences and elsewhere have an opportunity to explore promising research topics
by a hands-on experience with neural network simulators applied to
classi_cation and prediction problems ranging from bio-medical sciences to finance
and business.
Prerequisites:
Stat503 or CIS511 and undergraduate understanding of
probability, statistics and linear algebra.
Texts:
Haykin S. Neural Networks: A Comprehensive Foundation (2nd Edition), Prentice Hall, 1999, ISBN 0-13-273350-1 (required).
Bishop, C.M. Neural Networks
for Pattern Recognition, Oxford University Press, 1996, ISBN 0-19-853864-2(optional).
Topics: will be tailored
to interests of the participants. Content will include:
I.
Supervised and Unsupervised Neural Networks
1.
Multilayer Perceptrons
2.
Radial-Basis Function Networks
3.
Committee Machines
4.
Principal Components Analysis
II.
Selected Advanced Topics
1.
Self-Organizing Maps
2.
Information Theoretic Models
3.
Temporal Processing
4.
Dynamically Driven Recurrent Networks
5.
Applications
III.
Reading and research projects presentations.
Grading: Homework (30%), midterm exam (20%), reading/presenting assignments
(20%) and an individual research project (30%).
Late Policy and Academic Honesty: No late
submissions will be accepted. Discussing materials with fellow students is acceptable, but
programs, experiments and the reports must be done individually.