Seminar Details
| Date |
17-3-2010 |
| Time |
16:30 |
| Room/Location |
DISI-Sala Conferenze 3 piano |
| Title |
Multimedia Data Processing: Efficiency, Scalability, and Effectiveness. |
| Speaker |
K. Selcuk Candan, Professor |
| Affiliation |
Computer Science and Engineering School of Computing-IDSE (CIDSE) Arizona State University |
| Link |
http://www.public.asu.edu/~candan/index.htm
|
| Abstract |
Abstract: Media analysis and mining involve processing of large quantities of real-time and/or stored data and measurements. Data (whether captured in real-time through sensory measurements or processed, materialized, and stored for later use) are many times accurate only within a margin of error. Moreover, in many applications, the utility of a data element to a particular analysis task depends on the usage
context. The fundamental principles that govern the next generation of media/data analysis middleware must include data and operator imprecision, relevance of data to a particular analysis task, and the interest and expertise of the knowledge consumer. In this talk, I will discuss challenges for and opportunities in developing efficient and effective analysis middleware to support large scale data processing and decision making applications, where the data elements, metadata, and the operations on the data may be imprecise. As a
specific example, I will introduce and describe RankMR for processing ranked queries in batch-oriented cluster environments. In particular, I describe how RankMR adaptively samples data from "upstream" operators to help allocate resources in a work balanced and wasted-work avoiding manner for top-k join
processing. Experimental results show that the proposed sampling, data partitioning, and join processing strategies enable RankMR to return utlity ranked results with high confidence and low-overhead.
|
|
|
 |
