When: Wednesday, February 21, 2018
Where: PGH 563
Time: 11:00 AM

Cost-Delay Aware Web Routing Using Reinforcement Learning

Speaker: Gandhimathi Velusamy

We propose a Learning-Automata based solution to cut down the electricity bill for an enterprise, which has its data centers at multiple geographical locations. Our method reduces the energy charges caused by the power utilization of servers that handle web load. By exploiting the spatio-temporal variations in electricity prices, our method routes the web requests to servers that incur the lowest energy costs while maintaining optimal latency.

Bio:

Gandhimathi Velusamy is a fourth year PhD student working with Dr. Ricardo Lent from College of Technology and with Dr. Jaspal Subhlok for her research. Her interests are Computer networking, Software defined networking, Load balancing and Learning Automata.

Image Content Analysis in Indexed Captioned Searchable Videos

Speaker: Mohammad Rajiur Rahman

The project 'Indexed Captioned Searchable Videos' converts classroom lecture videos into topical segments to help students find a specific topic of interest quickly. The research explores use of visual similarity of the frames to improve creation of topical segments and find summarization of the video lecture in a single frame. Classroom lecture presentations contain texts and visual components like images, graphs etc. Along with texts, visual components can also be helpful in detecting topical segments. The idea behind it is: if two or more frames have similar visual components they are more likely describing same topic. Finding visual similarity of frames includes three steps: detecting visual component on a frame, finding component-wise visual similarity on different frames and then finally calculating frame-wise visual similarity. Component-wise similarity can be used to generate summarization of the video, where visual content from different frame can be put in a single frame to represent the content of the lecture.

Bio:

Mohammad Rajiur Rahman is a 2nd year PhD student and his advisers are Dr Jaspal Subhlok & Dr Shishir Shah. His research interest lies on studies related to Digital Image Processing, Machine Learning and Educational Technologies.