Homework Assignments

There will be three homework assignments in the class: one on self-organizing maps, and one on backpropagation, and one on game agents. These topics were selected because they illustrate the three main learning settings, i.e. unsupervised, supervised, and reinforcement-based learning; they also make it possible to compare different approaches in a competitive setting.

The assignments have two goals: (1) to obtain deeper understanding on how these architectures work and what their limitations are, and (2) to obtain practical experience in running these algorithms and to become familiar with the simulation programs, so that you can easily get started on the simulations for your project.

The assignments are somewhat open-ended in that you can spend quite a bit of time exploring the variations of the algorithms and different training regimes etc. The point is not to just get them done with minimal effort to get a good grade in the class, but to use the assignments as a starting point for in-depth exploration. You'll get more out of the assignments if you put more effort into them.

The assignments are turned in electronically to the TA using the Canvas system before midnight on the day they are due.

Sun Aug 3 22:58:25 CDT 2014