|Search Duke CS||
Undergraduate Research Poster Day 2014
Computer science undergraduates presented research projects April 21 during the fourth annual Computer Science Undergraduate Research Poster Day.
Brian Bullins, advised by Alex Hartemink, was selected to receive "Best Poster" recognition for his project, "Modeling Bias in DNase-seq Data for Improved Chromatin Occupancy Prediction."
Poster Day recognizes the achievements of participants in the C-SURF program and other Graduation with Distinction candidates. The requirements for graduation with distinction include maintaining a GPA of at least 3.0, completing a one year project with at least one semester of independent study, and presenting the results of the project to a committee of three faculty members. In addition, Poster Day is intended to generate interest and enthusiasm for undergraduate research among students.
The students below presented posters this year.
Project: Modeling Bias in DNase-seq Data for Improved Chromatin Occupancy Prediction
Advisor: Alexander Hartemink
Abstract: Whether or not a single gene is transcribed relies on a myriad of stochastic factors which may not be adequately described by the cell's genome alone. Understanding the connection between the occupancy of a cell's chromatin and the transcription of its genes would provide insight into the dynamic regulatory dependencies that control its internal transcription state, and so enhanced techniques for modeling chromatin state would be advantageous. In this thesis we consider improved methods of integrating DNase-seq data as input for a statistical inference model which outputs a nucleotide-resolution probability distribution of the genome's chromatin occupancy profile.In particular, we focus on some initial observations made concerning the probabilistic distribution of permitted cuts as part of the DNase-seq data, as well as extending the multivariate model so that it may account for sequence bias that is exhibited by the DNase I enzyme.
Project: Beat the Cheater: Computing Game-Theoretic Strategies for When to Kick a Gambler out of a Casino
Advisor: Vincent Conitzer
Abstract: Gambles in casinos are usually set up so that the casino makes a profit in expectations - as long as gamblers play honestly. However, some gamblers are able to cheat, reducing the casino's profit. How should the casino address this? A common strategy is to selectively kick gamblers out, possibly even without being sure that they were cheating. In this paper, we address the following question. Based solely on a gambler's track record, when is it optimal for the casino to kick the gambler out? Because cheaters will adapt to the casino's policy, this is a game-theoretic question. Specically, we model the problem as a Bayesian game in which the casino is a Stackelberg leader that can commit to a (possibly randomized) policy for when to kick gamblers out, and provide efficient algorithms for computing the optimal policy. Besides being potentially useful to casinos, we imagine that similar techniques could be useful for addressing related problems - for example, illegal trades in financial markets.
Project: Improving the Capabilities of JFLAP: Creating Effective User Interfaces in Learning for Theoretical Computer Science
Advisor: Susan Rodger
Abstract: JFLAP, the Java Formal Languages and Automata Package, is a software visualization tool for teaching various topics in theoretical computer science, including formal languages theory, grammars, and automata. Consisting of elements to represent formal languages of different classes (including regular, context-free, and recursively-enumerable), types of automata (including nondeterministic finite automata, pushdown automata, and Turing) machines, grammars, regular expressions, context-free and regular pumping lemmas, and Lindenmayer Systems (L-Systems), JFLAP is used internationally in various courses at the college level. The current version available to the public, JFLAP v7.0, is restrictive in its static codebase, not allowing for easy modification or additions, conceals the underlying formal definitions, and limits symbol naming conventions. To improve upon JFLAP's capabilities, much of the previous code has been redesigned, focusing on new class-based hierarchies to extend its flexibility, improving upon existing concepts to make the graphic user interface easier to interact with, and adding new functionality to broaden its capabilities. This reconstruction and extension has dealt with many of the issues facing the old version of JFLAP while maintaining similarity to avoid user confusion. To compare the two versions, an assessment survey was conducted regarding some of the new and modified capabilities. The survey indicated that the changes improve upon the old version and expand JFLAP's pedagogical capabilities.