Course Projects and Papers

During my time here at UW, I have worked on several course projects that despite being unrelated to my primary research, I am proud enough of them to showcase them on-line.

This paper was a survey of the current methods used in software verification for determining loop invariants. Given my background and interest in artificial intelligence, this was a truly interesting topic for me to explore. In the paper, I also discuss other applications of AI for finding loop invariants.

This project looked at using perceptrons (the most basic form of neural networks) for performing branch prediction. Starting from the work of Jimenez, et al. (see paper for the reference), we utilized a different and more realistic testing approach that gave mediocre performance to the branch predictor. To improve performance, we investigated the use of local history and caching.

This project explores a common scenario explored in real-time systems: the traffic light. A typical extension to a basic traffic light is to have it turn green whenever an ambulance approaches from a specific direction. In this project, I proposed a further extension that uses message propagation to improve the response rate of traffic lights in regards to approaching traffic lights. Furthermore, I developed three applets that demonstrate these scenarios.

This project investigated using the PlanetLab network for web anonymization. Anonymization is more than just encryption in that you also need to protect where you are sending from. We defined several anonymization scenarios and developed a system that allows the sender to control how much anonymization to be used.

This paper was a survey of the NLP problem of word sense diambiguation, i.e. determining the appropriate meaning of a word given its context. I give an overview of applications, resources, heuristics, and the current state-of-the-art techniques. Also included is a brief experiment involving the use a Naive Bayes classifier for performing lemma-independed pattern classification of WordNet senses.