Senior Thesis Talks

Middlebury College
Department of Computer Science Seminar

Senior Thesis Talks

Monday, May 7 at 12:30 PM:

Parallelizing Numerical Integration

Petar Mitrevski ‘07

Abstract: The ability to evaluate definite integrals is essential for virtually all fields of science. While integration has been known since antiquity, there are still functions whose anitderivative is not known, or computing it would be time-consuming. It is possible to approximate the integral of such functions using numerical integration. In this talk we will introduce some of the available numerical integration methods and compare them. Furthermore, we will suggest how to speed up these methods and increase their accuracy, using a cutting-edge technology known as parallel computing.

Monday, May 7 at 1:00 PM:

"Quantum Computation and Simulation"

Nikitas Stamatopoulos ‘07

Abstract: Quantum computation is a relatively new yet well researched topic. Given the rapidity that computing power increases with, it was foreseen that classical computers and classical computation would reach a maximum point after which speedup would slow down and eventually stall. The need for ever increasing computational power, along with the unique dynamics of quantum mechanics, gave birth to the field of quantum computation with the ultimate goal of creating a computer that would take advantage of the peculiarities of quantum mechanics. Even though a quantum computer is yet to be built in a complete form, the theory behind its operation has been developed and algorithms that would run on it have been devised.

We use the programming language QCL (Quantum Computation Language) to simulate on a classical computer the processes that would take place in a quantum computer. After going over important concepts in linear algebra, quantum mechanics and computer science, we delve into the world of quantum algorithms. We demonstrate the structure of QCL programming language and use it to simulate several well known quantum algorithms. Eventually we use Grover’s quantum search algorithm to implement a symmetric known plaintext attack in QCL.

Wednesday, May 9 at 12:30 PM:

“A Comparison of Image Segmentation Methods”

Anna Blasiak‘07

Abstract: The field of computer vision is concerned with extracting information from images. The task of image segmentation is a first step in many computer vision methods and serves to simplify the problem by grouping the pixels in the image in logical ways. Image segmentation is hard to clearly define because there are many levels of detail in an image and therefore many possible ways of meaningfully grouping pixels. Additionally, after choosing a definition for an optimal segmentation, there are many computational difficulties in finding such a segmentation. This talk with present three different ways to approach image segmentation, explain an efficient implementation for each approach, and show sample segmentations results.

Wednesday, May 9 at 1:00 PM:

"An Introduction to Robot SLAM"

Bradley Hiebert-Treuer ‘07

Abstract: SLAM is the problem of developing a method for robots to simultaneously localize and make maps of real world environments through the utilization of (imprecisely) observable landmarks positioned at fixed locations. This project introduces SLAM and describes three of the most effective solutions to the SLAM problem. The talk will present a full yet intuitive understanding of how the selected solutions function and why they are effective.

Friday, May 11 at 12:30 PM:

“To Block or Not to Block”

Kevin Chirls ‘07

Abstract: Synchronization is the problem of coordinating multiple events across time within a single system, such that the parallel events occur successfully without collisions. For example, without synchronization, two people could corrupt a shared file by modifying the file separately at the same time. This thesis explores two solutions to the problem of synchronization, the more traditional blocking synchronization, and then the newer innovative non-blocking synchronization.

Friday, May 11 at 1:00 PM:

"Landmark-Based Navigation Using the Expected Shortest Paths Algorithm"

Henry Roth ‘07

Abstract: Landmark-based navigation is a popular method for moving a robot around its environment. Landmarks are features of the environment that have been previously mapped and have a known relationship. But what happens when landmarks themselves can not be reliably detected? The Expected Shortest Paths algorithm gives us a strategy for dealing with the uncertainty of the real world.

Lunch will be provided at 12:15 p.m.

All are welcome to attend

This event is sponsored by the Computer Science Department.