Content-Type Coding

Speaker: Linqi Song, Ph.D. Candidate
Affiliation: UCLA - Advisor: Prof. Christina Fragouli

DEPARTMENT RESEARCH FORUM – SIGNALS & SYSTEMS

Abstract:  We observe that communication networks today are increasingly used to serve a fundamentally different traffic, that delivers type of content rather than specific messages. Content-type traffic pervades a host of applications today, ranging from search engines and recommender networks to newsfeeds and advertisement networks. Motivated by the aforementioned observation, we consider the coding problem where messages are clustered into content types. Our research first asks a novel question: if there are benefits in designing network and channel codes specifically tailored to content-type requests. We term this coding scheme content-type coding. In this talk, we will show through examples that in some cases we can have significant such benefits using content-type formulations. Our first example is to formalize a content-type coding problem in the index coding framework, where we propose algorithms to show that exponential gains are possible over message-specific coding scheme. A second example is given to show benefits in lossy networks: we consider a broadcast erasure channel with feedback setting where a source wants to send messages to two receivers. We quantitatively show that in this setting, content-type coding can achieve a larger capacity region than the message-specific coding scheme. Our third example shows that over a combination network, benefits of content-type coding over message specific coding tend to be the largest possible value, the type size (i.e., the number of messages within a type, this can be arbitrarily large in today’s Big Data applications), as the network becomes large enough.

Biography:  Linqi Song received the B.S. and M.S. degrees in Electronic Engineering, Tsinghua University, China. He is currently a PhD candidate under the supervision of Professor Christina Fragouli in the Electrical Engineering Department, UCLA. His research interests lie in the areas of index coding, network coding, algorithms and complexity, communications and networking.

Date/Time:
Date(s) - Jan 20, 2016
12:00 pm - 1:30 pm

Location:
E-IV Tesla Room #53-125
420 Westwood Plaza - 5th Flr., Los Angeles CA 90095