Sep 30, 2016
Lopata Hall, Room 101
Advisor: Brendan Juba
"A Polynmial Optimization Framework for Object Detection"
We introduce a new contour-based approach to shape based image segmentation. We make use of a global polynomial optimization framework to pose the problem of fitting specific shapes and introduce a learning based polynomial invariants framework for detecting them. One advantage of this method is that it doesn't require a clean image for training phase. The empirical results show that after rounds of learning, the algorithm is capable of detecting the desired shape in the contour of an image.
Advisor: Ben Moseley
"Online Scheduling for Average Flow in the DAG Model"
Online scheduling studies the problem of ordering the execution of a set of jobs as they arrive over time to some system. In this talk, the goal is to minimize the average flow time of the jobs - by far the most popular performance metric in online scheduling. This problem is well studied in many models, but we introduce the first results in the Directed-Acyclic-Graph (DAG) model of parallelism. In this model, each job corresponds to a DAG and may be processed on multiple machines at the same time. Most of the technical details will be abridged in the interest of time.