Intelligent Systems at Florida Tech

At Florida Institute of Technology, intelligent systems is one of the main areas of research in the Computer Sciences department. The focus in general is on (i) how to make computers more intelligent as well as (ii) how intelligence can change the ways we compute. Speci cally , one investigates algorithms that can help computers reason (constraint reasoning, spatio-temporal reasoning), learn (machine learning), and see (computer vision). Moreover, we examine how distributed intelligent agents can interact (distributed constraint reasoning and coordination). Our research also includes approaches on looking at how simplistic animal behavior can provide a novel way to solve problems (swarm intelligence). The ability to reason is fundamental to human intelligence for making decisions. The constraint reasoning group fo-cuses on spatio-temporal constraint reasoning. The group has developed new calculi for qualitative spatial reasoning that has applications in geographical information systems. The group has also developed techniques in detecting the culprit constraints in an unsatis able temporal reasoning problem. Recently, we launched a project on understanding creativity from a constraint reasoning viewpoint. The long term vision of this initiative is to develop an intelligent workbench for helping physicists in their creative activities. Another fundamental aspect of human intelligence is learning—the ability to generate new knowledge and adapt to the changing environment. The machine learning (data mining) group focuses on investigating techniques for anomaly detection and web personalization. Unlike the typical machine learning problem of building a classi er from training examples from two or more classes, the anomaly detection problem necessitates constructing a classi er from training examples from only one class—the " normal " class. The learned classi er is an anomaly detector that identi es and scores anomalies. For intrusion detection , anomaly detection has the potential of detecting novel attacks, which cannot be detected by identifying signatures of existing attacks. For device monitoring with time series data, we extract features , plot them, and generalize them into a sequence of " boxes, " which form a model for anomaly detection. For web personalization, we learn a user pro le from a user's bookmarks and use the pro le to re-rank results returned by a search engine so that the top results are closer to the user's interests. As eyes are windows to the world, the ability to interpret what we see is crucial in providing information for reasoning and learning. The computer vision group concentrates on object …