REU Site: CAAR: Combinatorics and Algorithms Applied to Real Problems

点击翻译按钮获取中文摘要

The project will recruit undergraduates to come to The University of Maryland at College Park where they will work on research projects. These projects are about combining theory and practice. For example, several are on machine learning which uses probability and statistics to train machines to find correlations in data, identify images, or play games. Other projects are in cryptography, where Mathematics is used to build secure systems. There will be a special effort to recruit students in the program from non-research schools and underrepresented groups. This program will give many of the students a chance to do research, which they otherwise would not have.This program will give students an idea of what graduate school is like in two ways: (1) their research projects are scaled down versions of PhD theses, and (2) there will be interaction with the students and current grad students.The students will be offered a variety of projects. We list some sample projects below:(1) Cryptography: Some next generation cryptosystems are based on the hardness of solving certain problems on lattices (rather than factoring). This project will explore implementations of known algorithms for solving these lattice problems as a way to test the security of the new systems.(2) Security: Side channel attacks are a way to attack a system by observing how much time or power (or other visible signs) the system uses. In the past such attacks have been used to find keys. In this project we develop and implement attacks that find out information about users' data, as well as develop and implement ways to prevent these attacks.(3) Allocation: How should a firm allocate its limited interviewing resources to select the optimal cohort of new employees from a large set of job applicants? Solving this problem seems to require interesting algorithms and machine learning techniques. In this project the students will implement programs for the problem and run them on real data.(4) Machine Learning/Image Recognition: A common problem in Machine Learning is to train a system to recognize an image, say of a dog. How to test such systems? Another problem in machine learning is to generate hard cases for an image-recognizer. In this project we will write programs that generate fake images that fool an image-recognizer, and we will use these fakes to improve the original image-recognizer.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.