|
Course description: Artificial neural networks, also called connectionist systems, are networks of relatively simple processing elements that mimic some of the properties of biological neurons. By studying these abstract nervous systems, researchers hope to improve our understanding of how real nervous systems function as well as to advance the field of artificial intelligence. This course provides and in-depth treatment of the central artificial neural network methods, with an emphasis on their computational capabilities. The objective of this course is to provide a enough knowledge of the basic elements of the field of artificial neural networks so that a student should be able to recognize problems for which artificial neural network methods are appropriate, implement such applications, critically evaluate literature in this field, and pursue self-directed further study.
Lecture: Tuesday & Thursday 2:30-3:45, CMPS 140
Prerequisites: Basic linear algegra, basic calculus, ability to write moderately complex computer programs (computer language unimportant). Students from psychology and neuroscience are welcome, provided they have some mathematical and computing skills. Contact the instructor if you have questions about your qualifications.
Credit: 3 units
Instructor: Andrew Barto, barto@cs.umass.edu, 545-2109. Office hours: Mon 1:30-2:30, Thurs 10:00-12:00, or by appointment, Computer Science 276
Required book: We will be using the textbook by Simon Haykin: "Neural Networks: A Comprehensive Foundation, Second Edition" Prentice Hall, 1999. The book is available at the textbook annex. Since it is quite expensive, you might try to find a used copy.
Assignments: There will be a set of exercises for each chapter. All exercises will be marked and returned to you. Some of these exercises will be programming projects. Each of these will require you to write a program, experiment with it, and answer a set of questions about its behavior. I will ask for volunteers to present their results in class. You will also be required to design a programming project of your own and hand it in at the end of the term: a short writeup describing it and the results. I will ask you to hand in your design ideas and work with with you to design a suitable project. You can use any computer to which you have access and any reasonably standard programming language (so I can help you debug your program if necessary).
Exams: There will be an closed-book in-class midterm and a clossed-book final exam during the exam period.
Grading: