Mark F. Rogers
Research Interests
"Algorithms are conceived in analytic purity in the high citadels of academic research,
heuristics are midwifed by expediency in the dark corners of the practitioner's lair."
- Dr. Fred Glover
Publications:
- M.F. Rogers, A. Howe and D. Whitley,
Looking for Shortcuts: Infeasible Search Analysis for Oversubscribed Scheduling Problems.
(ICAPS 2006) ps.gz, pdf
- Classifier assessment in protein function prediction
Researchers have noted that a classifier's performance statistics may be biased if the data used for
training and testing includes protein annotations derived through processes similar to the classifier.
To date little attention has been paid to the issue of bias in classifier performance assessments.
We are conducting a set of controlled experiments using a nearest-neighbor classifier based
on PSI-BLAST similarity scores.
Our results demonstrate that the source of GO annotations used to assess a protein function predictor can
have a significant influence on classifier accuracy.
Related articles:
- Alternative Splicing in A. thaliana
In this project we intend to develop classifiers that are able to detect alternative splicing in DNA sequences for A. thaliana.
In these plants, the most common form of alternative splicing is intron retention,
where introns are retained in mature mRNA under some conditions.
We are experimenting with classifiers such as Support Vector Machines (SVMs) based on Hidden
Markov Models (HMMs) and others based on Smith-Waterman alignment scores.
Related articles:
- Scheduling
Investigated the role of infeasible solutions in constrained
optimization problems (COPs). Many successful algorithms search both feasible and
infeasible solutions to achieve superior search performance.
Experimental results suggest that infeasible space adds a quantifiable improvement
over feasible space in some circumstances. Researchers hypothesize
two reasons for this behavior:
- COPs tend to have optima near the boundary between feasible and infeasible solutions
- Infeasible search may yield shortcuts to optima
Recent research confirms the presence of shortcuts, but we wish to learn how shortcuts
depend on specific problems and neighborhood operators. Our research examines infeasible
search with the following problems:
ROADEF Challenge 2003 ,
AFSCN Scheduling,
other EOS problems.
Resources:
- Poster for the 2005 ISTEC Colloquium pdf
- Poster for the 2006 CS Research Symposium (voted Best Masters Poster) pdf
- Poster for the 2006 ICAPS Doctoral Consortium pdf
- Consortium posters from CSU:
Mark Rogers,
Mark Roberts,
Andrew Sutton
Computing Theory:
