Computer Science Colloquium
Thursday, March 7, 2013
Davis Marksbury Building
University of Illinois at Urbana-Champaign
Collaborative Trustworthy Sensing in Cyber Networks
Abstract: Cyber network is a network of interconnected agents which collectively sense and claim factual knowledge and information about physical world. Many multi-agents systems in real world, such as social media networks, crowdsourcing systems and sensor networks, can be abstracted as multi-source sensing problem for cyber networks. Agents (e.g., information and data sources) in these systems often interact with each other, collaboratively or competitively, where one agent can be influenced by others in such an interactive environment. This leads to unequal degrees of trustworthiness among agents, making their claims noisy or even conflicting with one another. In this talk, I will present a novel idea of collaborative trustworthy sensing. It reveals the quality of information contributed by different agents, and study how the interdependence between different agents impact on agent trustworthiness in collaborative sensing and knowledge aggregation tasks. I will present a unified latent probabilistic model to connect the aforementioned latent factors to the quality of information and knowledge shared by interdependent agents. Efficient variational approach, which is guaranteed to converge to a stable solution, is derived to infer the most probable true values and interdependence structure underlying cyber network topology given the massive quality-deficient data collected from cyber networks. I will demonstrate that the proposed algorithm achieves competitive performance on aggregating knowledge and information from a big number of agents in real-world large-scale cyber network for a wide range of applications.
Bio: GuoJun Qi is a Ph.D. candidate in the Beckman Institute and the Department of Electrical and Computer Engineering in the University of Illinois at Urbana-Champaign. He has been awarded Microsoft Fellowship, IBM Fellowship and the best paper award at ACM Multimedia 2007. His work has appeared in the venues such as ICML, WWW, CVPR, ICDE, WSDM and SDM. His current research interests concentrate on integrating information and knowledge from multi-agent systems based on social and physical networks to support collective decision-making and problem-solving with the help of the big data in real world.