Quantifying the Value of Better Recommendations
by Neil Hunt (Netflix, USA), sponsored by Google
Moderator: Martin Ester
Recommender systems that are used to help users discover interesting content — for example: music or video — seek to improve the efficiency with which users engage with the content. How should we think about the value delivered? Is it about finding hidden gems that might otherwise not be discovered, or about shortening the time browsing to find something acceptable, or improving the appropriateness of a title within a specific search time, or something else? And how much value is delivered – are sophisticated systems worth the cost? Are real-time recommendations more useful than off-line precomputed recommendations, and if so, how much more is worth spending?
Dr. Neil Hunt is the Chief Product Officer of Netflix, which offers the world’s largest subscription service streaming TV shows and movies over the Internet. Neil’s team is responsible for the design and technology behind the Netflix service – the website, mobile, and Smart-TV experiences where members subscribe, pick movies and TV shows using personalized recommendations, search, and social suggestions, and stream them to any of over 1,000 different kinds of viewing experiences. The team is also responsible for the delivery pipeline, spanning encoding, storing, and serving tens of billions of hours of streamed content annually to more than 44 million Netflix members in 41 countries.
In addition to using data to help users discover great content to watch on Netflix, Neil is interested in personalized medicine, including the quest to pool data from cancer patients in a “Cancer Commons” — where big data techniques promise to deliver personalized treatment suggestions , by uncovering shared characteristics across cancer types and individual mutations to identifying how drugs designed for one cancer type may help an apparently different disease.
Prior to joining Netflix in 1999, Dr. Hunt worked from 1991 in various engineering and product roles at Pure Software and its successors Pure Atria and Rational Software building software testing products. Before that, Neil was engaged in research in computer vision and image processing at the University of Aberdeen, at Schlumberger Palo Alto Research Labs and Teleos Research.
Dr. Hunt holds a Ph.D. in Computer Science from the University of Aberdeen, U.K. and a Bachelor’s degree from the University of Durham, U.K.
Large Scale Machine Learning for Predictive Tasks
Keynote by Jeff Dean (Google Inc., USA), sponsored by Yahoo
Moderator: Michelle Zhou
Jeff Dean joined Google in 1999 and is currently a Senior Google Fellow in Google’s Knowledge group. He co-developed the MapReduce computational framework, and is a co-designer and co-implementor of heavily-used distributed storage systems, including BigTable and Spanner.
He co-designed and implemented five generations of Google’s crawling, indexing, and query serving systems, as well as major pieces of Google’s initial advertising and AdSense for Content systems. Most recently, he has been working on large-scale deep learning systems, with applications in speech recognition, computer vision, natural language processing, user action prediction, and user recommendations.
In this talk, Jeff will discuss work that he and his collaborators have done over the past few years in using very large deep neural networks to solve a variety of problems across many different domains. He will give a general overview of the types of approaches they have used, and discuss the parallelization strategies that they use for training and using these models. He will also discuss ways in which some of these models can be used for a variety of predictive tasks, and how they might be applied to personalization and recommendation systems.
Thoughts on the Future of Recommender Systems
by Hector Garcia-Molina (Stanford University, USA), sponsored by Pandora
Moderator: Alfred Kobsa
Hector Garcia-Molina is the Leonard Bosack and Sandra Lerner Professor in the Departments of Computer Science and Electrical Engineering at Stanford University, Stanford, California. He was the chairman of the Computer Science Department from January 2001 to December 2004.
From 1997 to 2001 he was a member the President’s Information Technology Advisory Committee (PITAC). From August 1994 to December 1997 he was the Director of the Computer Systems Laboratory at Stanford. From 1979 to 1991 he was on the faculty of the Computer Science Department at Princeton University, Princeton, New Jersey. His research interests include distributed computing systems, digital libraries and database systems. He received a BS in electrical engineering from the Instituto Tecnologico de Monterrey, Mexico, in 1974.
From Stanford University, Stanford, California, he received in 1975 a MS in electrical engineering and a PhD in computer science in 1979. He holds an honorary PhD from ETH Zurich (2007). Garcia-Molina is a Fellow of the Association for Computing Machinery and of the American Academy of Arts and Sciences; is a member of the National Academy of Engineering; received the 1999 ACM SIGMOD Innovations Award; is a Venture Advisor for Onset Ventures, is a member of the Board of Directors of Oracle, and is a member of the State Farm Technical Advisory Council.
In this talk, he will explain why as a user he is not happy with current recommender systems. Then he will discuss some research directions that might address some of the shortcomings. In making the case, he will discuss his experience with two systems he has been involved with, CourseRank and DataSift.
