Seminar by Vishwakarma Singh
Accurate and Efficient Similarity Search in Very High Dimensional Spaces
Vishwakarma Singh
University of California, Santa Barbara
Date: Monday, November 15th, 2010
Time: 3:15 PM
Venue: CS102.Abstract:
Similarity search is an integral part of many applications across various domains. One of the models of similarity search is to transform objects into vectors and then, design a multi-dimensional access method in vector space for search. Dimensions of the vectors for objects like images or documents can be in hundreds. Efficient and accurate search of similar objects in very high dimensions is vital for many applications. In this talk, I will introduce the similarity search model and briefly discuss why existing techniques fail to achieve all the required characteristics of a similarity search method for data of very high dimensions. I will discuss in detail a novel algorithm, SIMP, designed to perform well for data of dimensions as high as 256. I will also present comparative results with four existing techniques on very large datasets of very high dimensions to establish SIMP's performance claims.
About the speaker:
Vishwakarma Singh is a 5th year PhD. candidate at the Department of Computer Science, University of California, Santa Barbara. He received his B.Tech. degree with honors in Computer Science and Engineering from IT-BHU. Prior to joining PhD program, he worked with Oracle and Qwest. He is a IEEE member. He received UGC and SAIL scholarships for his undergraduate program.