Program Schedule: Friday, October 19 - Session 1
PhD Forum: HCI Applications
Location: South Ballroom
Natural Language Interfaces for Environment Control
Ana R. Chang, UC Berkeley
Supporting Data-Based Decision-Making for Caregivers through Embedded Capture and Access
Julie Kientz, Georgia Institute of Technology
PhD Forum: Algorithms
Location: Salon VI and VII
Automatic Bayesian Learning Methods
Jo-Anne Ting, University of Southern California
Mining Regional Knowledge in Spatial Datasets
Wei Ding, University of Houston
Slicing the Three-layer Architecture: A Semantic Foundation for Behavioural Specification
Michelle L. Crane, Queen’s University
New Investigators: Simulation and Performance Evaluation
Location: Salon I and II
A simulation system for ad hoc query-ready sensors
Lin Xiao (University of Illinois at Chicago), Sudeepta Musti (University of Illinois at Chicago), Aris Ouksel (University of Illinois at Chicago)
Sensor networks can be utilized in a wide spectrum of commercial, environmental, health care and military applications. This paper presents an integrated simulation platform we developed at the University of Illinois at Chicago to study and compare algorithm performances in the areas of localization, event handling, network topology control, routing, and query processing in various types of sensor networks. Our research strategy and results are also briefly outlined in this paper.
Performance Evaluation of ALLIANCES
Xinhua Yang (Colorado School of Mines), Xinhua Yang (Colorado School of Mines), Ed Krohne (Colorado School of Mines), Tracy Camp (Colorado School of Mines)
ALLIANCES is a MAC protocol that aims to resolve collisions. ALLIANCES can attain high throughput by exploiting the network diversity through the cooperation of nodes. In this paper, we analyze the results from the simulation in NS-2. After comparing the performance of ALLIANCES and IEEE 802.11b, we conclude that ALLIANCES offers increased throughput under high traffic load. Furthermore, the performance of ALLIANCES is improved further by extending the cooperative transmission epoch.
New Investigators: Artificial Intelligence
Location: Salon III
Simplifying Sketch Recognition UI Development
Tracy Hammond (Texas A&M University)
Sketch recognition systems are time-consuming to build and require signal-processing expertise if they are to handle the intricacies of each domain. Our goal is to enable user interface designers, who may not have expertise in sketch recognition, to be able to build these sketch systems. We have built GUILD to automatically generate sketch recognition UIs from hand-type LADDER descriptions, or from descriptions generated from a single example and system-generated near-miss examples.
Efficient Plan Recognition for Dynamic Multi-agent Teams
Gita Sukthankar (University of Central Florida), Katia Sycara (Carnegie Mellon University)
This paper addresses the problem of multi-agent plan recognition. We demonstrate the suitability of our new plan formalism for multi-agent plan recognition. From this representation, we extract local temporal dependencies that dramatically prune the hypothesis set of potentially-valid plans. Although multi-agent plan recognition is theoretically more computationally expensive than single agent plan recognition, we show that the presence of agent resource dependencies significantly reduces the set of potentially-valid plans.
New Investigators: Computing and Biological Applications
Location: Salon VIII
CIinBios: An Ontology-based System for Competitive Intelligence in Bioscience
Jiao Li (Department of Computer Science and Technology, Tsinghua University), Minlie Huang (Department of Computer Science and Technology, Tsinghua University), Xiaoyan Zhu (Department of Computer Science and Technology, Tsinghua University)
Bioscience becomes a strategic growing field for both academic institutions and industrial companies because of its profound impact on human health and clinical therapy. In this paper, we present an ontology-based system for competitive intelligence in bioscience (named as CIinBios), which supports decision making by searching bioscientific discoveries semantically and tracking new trends statistically. The experiments performed on 15,433,668 MEDLINE records yield evidence for the feasibility and validity of our system.
Jointly Learning Biological Interactions from Multiple Data Sources: Application to Pharmacogenomics
Xin Zhang (Arizona State University)
Pharmacogenomics studies the inherited variations in genes that are related to patient’s drug response. This field is very promising with anticipated outcome that drugs may one day be adapted to everyone’s own genetic features. We present joint learning frameworks and algorithms for inferring gene regulatory networks, predicting DNA-binding proteins from protein-DNA interactions, learning protein-protein interaction networks, and applying them to identify interactions between genes and drugs to select drug target.
Making the Future Web Accessible to People with Disabilities
Location: Narcissus and Orange Blossom
Presenters: Shawn Lawton Henry (W3C Web Accessibility Initiative (WAI))
It is vital that the Web is accessible for people with disabilities and older people with changing abilities, given its increasingly key role in education, employment, government, commerce, health care, recreation, and more. This presentation discusses the current state of Web accessibility and explores how we can each play a role in ensuring that the future Web enables greater participation in society instead of creating additional barriers.