CS 278: Evolution of Neurally Controlled Robots.


Instructor           Jaime Davila
Office              : ASH 204
Office Hrs       : Tuesdays and Thursdays, 1 - 2 PM
                              Monday and Wednesdays, 1:30 - 2:30,                          
                              or by appointment.
Phone Number : 413-559-5687
email               : jdavila at hampshire dot edu  (by far the best way to reach me)

This course is designed to give you an introduction to two important fields in artificial intelligence and machine learning: neural networks and genetic algorithms. In particular, we will ocncentrate on the intersection of these two fields. Emphasis will be placed on the design and implementation of course projects.

There will be a fair amount of material covered during the first few weeks of the semester (enough to enable students to start working on their own projects). After that, we will spend a good portion of our meetings discussing the progress and pitfalls of our own projects, coming with solutions and next steps in a collaborative way. This means that students will need to be deeply engaged with the course material on a weekly basis.

Although several neural network, genetic algorithm, and visrtual environment programs will be made available to students, developing individual projects will require making modifications to (and/or interfaces for) these programs. For this reason, students should have programming experience before registering for this course. Students not sure about their level of programming expertize are encouraged to ask the instructor about their fit for this course.


Your evaluation for this course will be based on 1) your development of a NN/GA experiment, and 2) class participation.

Some online resources.

1) NN Frequently asked questions (FAQ)

2) GA Frequently asked questions (FAQ)

3) Documentation on class-provided virtual worlds.

4) Documentation on GENDALC.

5) SNNS home page (including online manual).

5b) Documentation on SpikeSNNS, which extends SNNS to include spiking neurons.

6) A quick tutorial on GANN created by the ANNEvolve group.

7) Documentation on Remote Procedure Calls (RPC).

Below is an outline of the topics we will discuss this semester. All of the basic material will be presented in class, but no textbook will be assigned. For this reason, attendance to all classes is HIGHLY recommended.

Bear in my, also, that some topics will be introduced as needed based on the conversations we will be having in class. Whenever we do, I will update this syllabus to reflect the topics covered.

Week Dates Topics
1 Sept. 4 Course outline.
What you should expect of the course.
Office hours, etc.
Introduction to the course.
First look at virtual worlds.
2 Sept. 9, 11 Crash course on neural networks.
3 Sept. 16, 18 Crash course on genetic algorithms.
4 Sept. 23, 25 The GENDALC GANN system.
5 Oct. 2 Creating hook-ups to virtual environments.
Levels of abstraction.
6 Oct. 7, 9 More on GENDALC and SNNS.
7 Oct. 16 Student project proposals.
8 Oct 21, 23 More on neural networks.
9 Oct. 28, 30 Progress reports.
10 Nov. 4, 6 Progress reports.
11 Nov. 11, 13 Updated project proposals.
12 Nov. 18, 20 Result validation.
13 Nov. 25 Progress reports.
14 Dec. 2, 4 Progress reports.
15 Dec. 9 Class presentations.