CS 152 Projects
Project Requirements
- The project contributes about half of your grade.
- It can be done individually, or in teams of two.
- If done as a team, double the ambition level is expected.
Action Items
- Decide whether you are going to do an individual or team project.
- If a team project, identify the team member.
- If you want to do a team project, but don’t have a good idea for a co-member, let me know and I will act as matchmaker.
Action Items
- Let me know by email as soon as possible which you intend to do, but no later than next Tuesday, 10/26.
- To get started with your investigation, choose and read one or more sources from:
- web pages (perhaps start with the list of links: http://www.cs.hmc.edu/courses/1999/fall/cs152/links.html)
- books and journals in the library
Action Items
- During class, week of 11/2, give a brief (say 10 mins.) presentation on a source you read, followed by an initial proposal.
- Establish a home page for your project and send me the URL.
- By 11/11, have a written proposal for your project installed on your home page.
- Check with me for final approval of your proposal.
Project Categories
- Demonstration programs or applets, that demonstrate some key concept (must be documented with theory and user info as well).
- Research paper that expounds on concepts, contrasts approaches.
- Experiments, either original or replicative, that verify previous work.
Constraints and Allowances
- Must be in the general field of neural networks, soft computing, artificial life.
- Must show new learning and understanding on your part.
- Need not be original (in the sense of new, publishable, knowledge)
Advice
- Try to assess feasibility before settling on topic.
- Don’t make it too ambitious (e.g. playing some complicated game).
- Don’t make it too “off the wall” (e.g. training a network to do something that cannot be accurately specified).
- Don’t try to predict the stock market.
Project Reporting Schedule
- (11/18 and 11/23) Oral, in-class, progress reports
- Week before and of finals: Final progress reports and demonstrations
Possible Project Areas
- Projects involving motion, robotics, kinematics
- Training a simulated unicycle rider
- Projects involving speech, sound, music
- Projects involving vision
- Implementations of known algorithms, such as variable-learning rate methods, conjugate gradient, Levenberg-Marquardt, etc.
Possible Project Areas
- Networks capable of growing or shrinking
- Extracting explanations from a network
- Improved weight initialization for backprop networks; use of simulated annealing to avoid local minima
Possible Project Areas
- Replicating classical experiments, such as NetTalk, Widrow’s truck-backer or pole-balancer, etc.
PPT Slide