::::::::::::: Courses :::::::::::::

Automatic Web Service Composition (08/2005 - Now)
(Directed by Dr. Prashant Doshi, major professor)

Paper 01: Composing Nested Web Processes using HSMDP
Haibo Zhao, Prashant Doshi, AAAI 2006 Workshop on AI-Driven Technologies for SOC(AAAI06, Boston)

Paper 02: A Hierarchical Framwork for Composing Nested Web Processes
Haibo Zhao, Prashant Doshi, International Conference of Service Oriented Computing (ICSOC06, Chicago)

Automatic Web Service Composition has been a big challenge in the Web services research community; we propose a decision theoretical planning approach (MDP/SMDP) to handle this problem. The primitive experimental results of our work are very encouraging and we observe that our approach outperforms some other approaches in terms of both the optimality of the process and the scalability. Our future work will focus on the parallel activity support and scalability improvement. We strive to provide an end-to-end solution for large-scale Web process composition problems.

LFY - A Google-based People Search Engine ( 01/2006 - 05/2006)
(with Haibao Tang)

Project Proposal
Project Report
Project Demo: http://www.haibozhao.com:8080/mlProject

Given a person's name, we can search for it in GOOGLE and get a huge number of results, but these results are normally not about the same person because so many people have the same name. It is annoying and difficult to identify what web pages are really about the person you are looking for.

Bayesien learning is utilized by our project to classify these results into groups like Athletic, Professor, Business Man, Office Worker, Government Employee, Actor, Dancer, and Singer.......

Cuadro: A semantic photo sharing community ( 08/2005 - 12/2005)
(with Pablo N. Mendes)

Project Statement

Cuadro is a semantic web based website to store, sort, search and share your photos online. By cuadro, uses can easily organize and share their huge number of photos with their family and friends, or the whole world if they want. Cuadro provides an integrated environment to annotate photos by generating semantic metadata of photos. Semantics is introduced into this application so that we can better describe all these photos and organize them by relationships. This means much more than just holding a digital image in a server and showing it on a web browser for users. Cuadro intuitive photo sharing system helps the user to describe the meaning of the photo contents, allowing for organizing and searching pictures based on the meaning of its contents, not only on the name the user provided

The system enables the association of photo descriptions with a computer interpretable representation capable of expressing the conceptual relations between entities in the real world. By navigating through the conceptual relations, the system can automate basic steps on picture description, sharing and interlinking, bringing live the real community.

Ontology Visualization (01/2005 - 05/2005)
(with Ravi Pavagada, Justin Milligan, Joe Dawley)

Project Statement

This tool is desinged for end users who are interested in viewing specific regions in the ontology. The tool would have a class tree view and property tree view as in protégé. This tool would be very useful for the end users who are interested in viewing specific regions in the ontology. The user will be viewing the condensed instance graph initially. The tool would have a class tree view and property tree view as in protégé. The user would be able to choose a node and or property (relationship) from the instance graph either from the instance and property view or by other ways of providing input (could be a dialog box or through a select option in popup menu when right-clicked on a node or property). The user would be able to view all paths that the selected property was involved between any two nodes selected. Hence, the tool would help to find out the semantic associations between any two entities.

MDGA - Motif Discovery Using Genetic Algorithms (08/2004 - 12/2004)
(with Dongsheng Che, Yinglei Song)

Project Report

In this project, we proposed an approach that can be used to computationally detect the binding site motifs on the upstream regions of genes. Based on the generic framework of a genetic algorithm, the approach explores the search space comprised of all possible combinations of starting locations of the binding site motifs in all the target sequences with an evolving population. Individuals in the population may stochastically participate in the evolution procedure which includes crossovers and mutations. Our experiments demonstrated that, compared with other approaches that use the Gibbs sampling algorithm, an improved level of prediction accuracy was observed using our approach. More importantly, the running time needed in our approach is independent of the length of target sequences and thus can be significantly reduced when the target sequences become very long.

Realtime System Scheduling (Restricted EDF) (12/2004 - 5/2006)
(RA directed by Dr. Shelby Funk)

Project Statement

Normally, a real-time OS needs to deal with both real-time jobs and background jobs. Different strategies (FF, LF, BF, WF) can be applied in R-EDF to schedule real-time jobs; background jobs can also be scheduled in many different ways. Different scheduling methods will give us very different response time for background jobs. So, the objective of this project is to find out the best strategy which gives us the shortest response time of background jobs without missing any real-time deadline ......