As part of CS150 Machine Learning,
students are expected to pick a particular topic,
study it in greater depth,
and report to the class on that topic.
This requirement serves several purposes
and accomplishes several goals.
it provides opportunities for students to practice
research and presentation skills.
Presentations will be a recurring component of life after college
and most problems will require some amount of research.
- Select a topic and have it approved.
You should start by selecting a topic of interest to you from the list below.
You must obtain approval for your topic at least two weeks prior to your scheduled presentation.
If you are selecting a topic from the list below,
first check the Eureka schedule to make sure none of your peers have already selected that topic.
Then send me an email with the topic you would like to explore.
You may also propose a topic that is not on the list,
but in this case you must obtain approval three weeks in advance.
- Delve into the topic via the research literature.
Having settled on an approved topic, locate several relevant sources.
I suggest you aim for four papers that you will read and draw upon.
At a minimum, you must work extensively from two sources beyond your textbook.
- Write a short research report.
Having learned about work on the topic in question,
write a report of the work that has been done.
Your report must be formatted in a research conference format
and may be no more than two pages (including references).
Your goal for this report should be to interest the reader in the topic
and describe some of the published work that addresses the topic.
- Prepare and deliver a short presentation.
You will make a 10-15 minute presentation to your peers.
As with the report, your goal should be to provide enough understanding
such that members of the audience can become interested in the topic
and explore it further on their own.
The following list of suggestions
may help if you are having trouble selecting a topic.
Note, in all cases, I expect you go beyond the treatment presented in the textbook.
You must read and digest at least two sources form the machine learning research literature.
Discretization of continuous attributes (Greg)
Recurrent artificial neural networks (Daniel)
- Support vector machines
- Boosting and bagging
Bayesian belief networks (Casey)
- Radial basis functions
Diversity in genetic algorithms (Austin)
- Temporal difference learning
Unsupervised learning (Kelly)
Efficient k-Nearest-Neighbor (Lewis)
Submission and mechanics
Your presentation dates have been assigned to you.
You may arrange with a peer to switch dates if the switch is mutually agreed upon.
In the case of a switch, one of the presenters must email me
and copy the other presenter;
your email must contain all of the information relevant to the switch.
Submit to Eureka your report.
If you choose to use presentation software,
also submit the slides from your talk.
If you perform a demonstration of some software,
include the code and instructions for running it.