UMBC ebiquity

Masters Thesis Research Proposal: Amey Sane and Nikhil Puranik

Speaker: Amey Sane

Start: Tuesday, September 27, 2011, 10:30AM

End: Tuesday, September 27, 2011, 11:30AM

Location: ITE 325-B

Abstract: In this week's lab meeting we will have Amey Sane and Nikhil Puranik talk about the research they will be pursuing for their Masters Thesis.

Amey Sane will talk about "Context-aware Framework for modeling and prediction of User activities". His research work is part of ongoing Platys project. This project spans over the areas like mobile computing, Context-aware computing, security and privacy. My thesis work will be focused in the area of mobile-computing and Context-aware computing. In today's world smart devices like mobile phones can effectively be exploited for use, not only as a mode of contact through voice dialing but much more than that, by making use of its features like Bluetooth , Wi-Fi capability and also through its in-built sensor capability. All these can prove to be effective in getting a notion of user context in terms of his current conceptual location (at home, at work or elsewhere) and activity(working, walking, attending meeting, exercising). Our research will focus on building a Hidden Markov Model for modeling and predicting User Activities. Also, we will eventually focus on building a generic prototype model for modeling users with different profile(Student profile, Working person's Profile, Lecture's profile). Learn more about Platys project here.

Nikhil Puranik will discuss 'A general framework for using data specialists for classification of data'. Given a collection of items, the framework will predict what is the label to which the items belong to? For example, given a column of data with all SSNs, the framework will come up with a label of 'SSN' for this column with a particular confidence. The framework will comprise of creating agents who will individually examine given data sets and come up with a probability of that set belonging to a particular class(concept) with its confidence measure and the decision on the class label will be made on the maximum confidence level. This framework will be based on the Blackboard Architecture in AI.

Host: Tim Finin

,