SICSA DVF Jimmy Lin, Associate Professor at the University of Maryland

Loading Map....

Date(s) - 22/06/2015
11:00 am - 1:30 pm

University of Edinburgh Informatics Forum

Professor Jimmy Lin, University of Maryland will be visiting the University of Edinburgh to give a talk on ‘Building Effective and Efficient Information Retrieval Systems’.

Machine learning has become the tool of choice for tackling challenges in a variety of domains, including information retrieval. However, most approaches focus exclusively on effectiveness—that is, the quality of system output. Yet, real-world production systems need to search billions of documents in tens of milliseconds, which means that techniques also need to be efficient (i.e., fast).

In this talk, I will discuss two approaches to building more effective and efficient information retrieval systems. The first is to directly learn ranking functions that are inherently more efficient—a thread of research dubbed “learning to efficiently rank”. The second is through architectural optimizations that take advantage of modern processor architectures—by paying attention to low-level details such as cache misses and branch mispredicts. The combination of both approaches, in essence, allow us to “have our cake and eat it too” in building systems that are both fast and good.

Jimmy Lin is an Associate Professor in the College of Information Studies (The iSchool) at the University of Maryland, with a joint appointment in the Institute for Advanced Computer Studies (UMIACS) and an affiliate appointment in the Department of Computer Science. He graduated with a Ph.D. in Electrical Engineering and Computer Science from MIT in 2004. Lin’s research lies at the intersection of information retrieval and natural language processing; his current work focuses on large-scale distributed algorithms and infrastructure for data analytics. From 2010-2012, Lin spent an extended sabbatical at Twitter, where he worked on services designed to surface relevant content to users and analytics infrastructure to support data science.

This entry was posted in .