Dec 09
Posted on

Probabilistic Inference in Social Networks

Constantinos Raftopoulos
Professor of Information Technology
School Of Liberal Arts and Sciences

When: Monday, March 7, 15:00 – 15:50

Where: Deree Faculty Lounge

Organized by: Faculty Research Seminars 2015-16 Series

 


Quantifying semantics has been a challenge in many disciplines ranging from social sciences to artificial intelligence, finance and education. In the emerging field of data mining in particular, the challenge takes the form of the “semantic gap”: the lack of coincidence that is, between the information that one can extract from the web and the interpretation that the same data have for a user in a given situation. In this talk we will discuss how the computational permissiveness of probability theory can address the issue. A novel system similar to Google Image Search will be presented as a paradigm where users ask queries – lists of keywords – and they choose or not some of the returning images. Users implicitly relate the images they choose to the keywords asked for these images but users also implicitly relate keywords to each other regardless of the chosen images. A probabilistic inference is realized for each image, transparently annotating/representing the image. The same model is used to define/quantify pair wise relevance between keywords. A semantic distance is thus probabilistically defined and used to cluster, users, images and keywords. The method has certain theoretical and practical advantages when compared to other state of the art methods.

This work has been published in IEEE Transactions in Knowledge and Data Engineering Journal, the #1 Ranked venue in Data mining worldwide, according to Google Scholar Metrics.


Constantine A. Raftopoulos, ED, PhD, Professor I: Ptychion in Mathematics (B.Sc, UOA), National and Kapodistrian University of Athens (1994), Master of Science (M.Sc, UCLA) in Computer Science, University of California at Los Angeles (1996), Doctorate (PhD, NTUA,) National Technical University Athens and Terminal Degree of Engineer (ED, UCLA), University of California at Los Angeles (2002). Currently at the rank of Professor I, he has been serving in the department of Information Technology at Deree College since 2015, teaching computer science related courses. He is an active researcher in the fields of data mining, computer vision, machine learning, pattern recognition, computer networks.