CS 294-4: Connectionist and Neural Computation

CCN: 24932

Web Page = http://http.icsi.berkeley.edu/~shastri/Courses/index.html

Tu and Th 12:30 - 2:00; 310 SODA


Jerome Feldman
email: jfeldman@icsi.berkeley.edu
office hour: Monday 1-2pm, 731 SODA or by appointment

Lokendra Shastri
email: shastri@icsi.berkeley.edu
office hour: Thursday 2-3pm, 731 SODA or by appointment


With over 100 billion parallel computing elements, the brain's capacity for encoding, communicating, and processing information is remarkable. But the brain is also extremely limited: neurons are slow devices, their computation appears relatively undifferentiated with respect to the needs of symbolic computation, and they communicate via messages that encode only a few bits of information. Despite these limitations, our brain is adept at performing certain types of symbolic processing and learning tasks with remarkable ease (e.g., learning and understanding language). In this course we will study how biological systems can perform such an impressive computational feats and how the resulting insights can be applied in engineering as well as science. The standpoint will be computer science, but will employ results and methods from several disciplines.

Textbooks ... Lecture Schedule ... Grading ... Reserved Readings


Prerequisite for the course is graduate standing in any field OR completion of the course: CS188: Artificial Intelligence OR permission of the instructors. The course will involve considerable use of simulation packages and some math, but programming will not be required. The required term project can, of course, be programming intensive.

Assignments ... Term Projects

Some useful resources

The Digital Anatomist Interactive Brain Atlas

Reinforcement Learning On-line Tutorial

Temporal Difference Simulation by Doug Rohde

More to follow