CSCE 990 - Support Vector Machines (Spring 2006) Topic Survey/Implementation Ideas (+ some possible research projects)
For a course project, more breadth is expected from a report (literature
reivew) than in an implementation. In both cases, a detailed description
of different approaches is expected, as well as a careful comparison
and a critical assessment. Contact me for more information on any of these.
- A study of kernels/optimization techniques for the
multiple-instance learning model, where each labeled example
consists of several instances from the instance space, and we don't
know which instance(s) is/are responsible for the label (report, maybe
implementation; could become a research project)
- The effect of approximate kernel computation on SVM optimization
(research project)
- Parallelizing kernel computation or SVM optimization
(implementation)
- Applying SVMs to information retrieval
(report or implementation; could become a research project)
- Applying SVMs to biological sequence analysis, phylogenetic analysis, or
other bioinformatics application (report or implementation;
could become a research project)
-
Kernel principal Component Analysis (KPCA) for feature selection (report or
implementation)
-
SVM-based automated computer system intrusion detection, i.e.
automatically identifing when a person has cracked into a computer system
(report or implementation)
-
Using an SVM to learn to classify web pages or usenet news (report or implementation)
-
Connecting the techniques of boosting and bagging to margin maximization
Back
Return to the CSCE 990 Home Page
Last modified 16 August 2011; please report problems to
sscott AT cse.