I had the pleasure of attending presentations by the top 3 undergraduate and top 3 graduate students who participated in the Student Research Competition at Grace Hopper. The posters of the 11 graduate entries and 5 undergraduate entries were evaluated on Wednesday evening, and the top 3 in each category went on to present their work on Thursday afternoon. All of the work is fascinating, and I know the judges had a difficult time deciding amongst the students.
Graduate Entries:
Computer Aided Insights Into the Biomechanics of Dinosaurs, presented by Zartasha Mustansar, University of Manchester, UK
In this project that brings together computer science an paleontology, Zartasha is using a number of computer based tools to gain insights into paleontology, based on simulation and biomechanics principles. She has used a range of methods including X-ray micro-tomography, meshes and smoothing, modling, and VR visualization.
Dynamic NBTI Management in Multicore Processor, presented by Taniya Siddiqua, University of Virginia
Taniya described two techniques that address the guardbanding practice used in to address NBTI problems at the chip level (I confess, though I am a software testing person, and generally leave it to the hardware folks to take care of that end of things, I found her presentation to be very clear and accessible). Taniya has shown that combining two techniques, an instruction scheduling policy and circuit-level "power gating", she is able to cut in half the size of the guardband that is necessary.
Handover Optimization in Fourth-Generation Heterogeneous Wireless Networks, presented by Xiaohuan Yan, Monash University
As 4G networks become more prevalent, the issue of handover between the 3G local network, WANs, and Wi-Max becomes increasingly important. Xiaohuan is exploring the questions involved in determining whether a handover needs to be made and exactly when that handover should be made. The goal is to make it possible for users to take advantage of the higher bandwidth and lower cost of 4G networks, without any loss of reliability as handover takes place.
Undergraduate Entries:
Presenting Clinical Survival Probability Charts on Mobile Devices, presented by Nan Meng, Winona State University
Nan has worked to develop a new interface that allows clinicians to access the web based Lung Cancer Survivability Prediction Tool from mobile devices. The goal of the clinicians is to be able to talk clearly with patients about the relative benefits of different treatment options (surgery, radiation, chemotherapy). When this tool is ported to an iPhone, a patient can easily see what the predicted life expectancy outcomes are and discuss this with their doctor. The two challenges Nan addressed were the speed of data transfer and the small display size of the mobile device. She presented nice solutions to these problems and also used the touch screen of the iPhone to allow patients or clinicians to pull up data from any portion of the graph that is displayed.
Recommendation-based Query Relaxation via Space Partitioning and Mapping Functions, Manasi Vartak, Worcester Polytechnic Institute
Manasi has tackled a very difficult problem, that of improving the outcome of queries without the extensive trial-and-error process that many of us go through when trying to find the best airfares, hotel deals, etc. After reviewing the approaches used by many other researchers in this area, Manashi used an approach that involves space partitioning and mapping functions that allows her to incrementally relax queries. Her work was extremely well presented and she had a fabulous command of the material, way beyond what I am used to from undergraduates!
Towards Emotionally Intelligent Machines: A Comprehensive Mood Classification System, Lucy Vasserman, Pomona College
Lucy’s work is in the area of affective computing, part of addressing the question of whether we can eventually build machines that have emotional intelligence. Lucy build a system that, trained on blog text, can identify the mood conveyed in the text. She based her work on Naive Bayes classification methods. I tend to think more about language generation than I do about language understanding, but I presume that once we have systems that can effectively identify mood in text, the next step will be to use the key elements to automatically generate text which conveys a particular mood, and do so in a way that is convincing to humans.
I was very impressed by the poise of all 6 students, as well as by the quality of their research and the content of their presentations!
Join the Discussion (0)
Become a Member or Sign In to Post a Comment