Lecture 1 - August 26, 1997
There are easily a dozen significantly different approaches to studying
Connectionist and Neural Computation, characterized by the original
field of the authors. Our background is Artificial Intelligence and
Parallel Computation and the course will reflect that. As always in advanced
courses, there will be considerable overlap with our ongoing research
project on connectionist models of language and thought: the
Neural Theory of Lanaguage Project
The three
recommended texts are by people with backgrounds in:
- Developmental Psychology - Plunkett and Elman
- Engineering - Ballard
- Statistics - Bishop
The reserved readings contain books that approach the subject from
several additional perspectives:
- Cognitive Psychology - Rumelhart and McClelland
- Mathematics - Rojas
- Physics - Hertz, Krogh and Palmer
- Theoretical Computer Science - Valiant
- Neuroscience - Churchland
There are many journals and conferences covering the field, but
the strongest are the
journal
Neural Computation and the two annual conferences: the
Cognitive Society Conference and the
Neural Information Processing
Society (NIPS) Conference. The Anderson book on reserve collects
many historically important papers. A good strategy for approaching
any new field is to scan the recent journals and conference proceedings.
There is an archive of pre-prints at Ohio State:
ftp.cis.ohio-state.edu/neuroprose.
We are also assembling a collection of web pages of other courses on
Connectionist and Neural Computation Other Courses.