Brief course description
CNS/Bi/Ph/CS 187. Neural Computation. 9 units (3-0-6); first term.
(Note that CNS 187 is a direct descendent of the former CNS 185,
which is no longer taught.) This course investigates computation
by neurons. Of primary concern are models of neural computation and
their neurological substrate, as well as the physics of collective
computation. Thus, neurobiology is used as a motivating factor to
introduce the relevant algorithms. Topics include rate-code neural
networks, their differential equations and equivalent circuits;
stochastic models and their energy functions; associative memory;
supervised and unsupervised learning; development; spike-based
computing; single-cell computation; error and noise tolerance.
This course is rather interdisciplinary as it attempts to pull
together topics in biology, physics, mathematics, and computer
science. However, lack of background in one or more of these areas
need not prevent you from succeeding in the course. We do require
some basics: familiarity with computer programming in a high level
language (ideally MATLAB), and knowledge of basic calculus,
probability, linear algebra, and some familiarity with differential
Useful Matlab references:
This course has three required textbooks,
Introduction to the
Theory of Neural Computation
by A. Hertz, A. Krogh, and R. Palmer,
Information Theory, Inference and Learning Algorithms
, by David MacKay, and
Spiking Neuron Models by W. Gerstner and W. Kistler.
All these books are available in the bookstore. MacKay's book is
also available online.
Although these books will be the main references, additional books and primary
source papers will be useful on a lecture by lecture basis.
Reference material will be available as appropriate.
The following books are also highly recommended:
Arbib, M., The Handbook of Brain Theory and Neural Networks
P. Dayan and L.F. Abbott, Theoretical Neuroscience
Maass, W., and Bishop, C., Pulsed Neural Networks
Anderson, J., Neurocomputing Foundations of Research
Ballard, D., Introduction to Natural Computation
Bishop, C., Neural Networks for Pattern Recognition
Kandel, E. Schwartz, J. and Jessell T.,
Principles of Neural Science
Koch, C., Biophysics of Computation
Minsky, M., Computation: Finite and Infinite Machines
Nicholls, J., From Neuron to Brain A Cellular & Molecular Approach to the Function of the Nervous System
PDP Research Group Staff, Parallel Distributed Processing Explorations in the Microstructure of Cognition, Vol. 1: Foundations
Rieke, F., Warland, D., de Ruyter van Steveninck, R. and Bialek, W.,
Spikes; Exploring the Neural Code
Rumelhart, D., Parallel Distributed Processing Explorations in the Microstructure of Cognition, Vol. 2: Psychological & Biological Models
Shepard, G., The Synaptic Organization of the Brain
Strogatz, S., Nonlinear Dynamics and Chaos
Other references that may be helpful.
Instructors, teaching assistants, and web page
CNS 187 is taught by Professor Erik Winfree.
If you have a question for the TAs, you are strongly encouraged to use
of the cns187 e-mail address firstname.lastname@example.org (as
opposed to the telephone or personal email). The TAs will hold office hours once per
week -- see the main class page. During their office hours, the TAs
will be available for assistance on any topic covered in the class but
will not be available for assistance regarding computer
programming. You are advised to regularly check the class webpage at
http://www.dna.caltech.edu/cns187; it will continually be
updated with homeworks, syllabus reviews, etc.
http://www.dna.caltech.edu/cns187 , so be sure to check it
Questions about homeworks or administrative details can be sent to the
email address email@example.com. This goes to the TAs.
There will also be a mailing list including the addresses of all the
students in the class,
firstname.lastname@example.org. Anything sent to it will be
automatically broadcast to everyone. The TAs will send homework
clarifications (if necessary) to this mailing list, so check your
There will be an assignment due each week covering the material taught
in the lectures and covered in the references. It is important to
read the references for a better understanding of the lectures and
assignments. However, do not consult materials that directly
provides a solution to the problem: every problem should can and
should be solved using only the information provided in lectures
(except where specifically indicated). For example, do not consult
problem sets or solutions sets from previous years.
The homework sets will include computer simulations in
MATLAB, mathematical derivations and analysis. Collaboration is
permitted for understanding and discussing all problem sets, but each
student must write their own submission in their own words and
understand everything they hand in. You should also write your
own matlab codes. Names of collaborators must be
given on the homework. This will not affect grading.
Problem set grading and lateness policy
Problem sets will be given a grade from 0 thru 100; a failing grade
(F) will be given for inexcusably poor or absent problem sets.
Decisions of the TA's regarding problem set performance will be
final. Answers are expected to be clear, concise, and thorough; if a
TA has questions about the answer your provide, you may be required to
speak to him for full credit (this will be clearly marked).
You may consult textbooks and web material about fundamental concepts
in neural network, but you may not consult such material if it
provides a solution to problem identical or nearly identical to the
one in the CNS 187 homework set. Furthermore, you may not consult
solution sets from previous years, or homework sets and/or solutions by
students who took the course in previous years.
Homeworks will be posted on the course webpage on Thursdays and are
due the following Friday before 5:00pm in the TA's mailboxes unless
otherwise noted. The class mailbox is located on the first floor of
Moore. Problem sets handed in late, but before 5pm on Monday will be
penalized by 20%. Problem sets will not be accepted after this time,
and therefore given an (F), as the solutions will have been posted on
the web by Tuesday morning. One free extension to Monday will be
granted to each student (don't use it up needlessly!); a longer
extension requires permission from the instructor. Further extensions
will only be granted for extreme circumstances (e.g. serious medical
or family issues). Any non-emergency extensions must be cleared before
the due date with the instructor or one of the TA's.
There will not be a midterm nor a final exam.
Course grading policy
The course will be graded on the basis of the regularly assigned
Computer programming and Web browsing are integral parts of the
course. Matlab is available on the ITS and UGCS computer clusters,
among others. If you need an account on the UGCS machines, use the UGCS/CS
Computer Account Request Form.