Big data makes it possible in principle to learn very rich probabilistic models, but inference in them is prohibitively expensive. Since inference is typically a subroutine of learning, in practice learning such models is very hard. Sum-product networks (SPNs) are a new model class that squares this circle by providing maximum flexibility while guaranteeing tractability. In contrast to Bayesian networks and Markov random fields, SPNs can remain tractable even in the absence of conditional independence. SPNs are defined recursively: an SPN is either a univariate distribution, a product of SPNs over disjoint variables, or a weighted sum of SPNs over the same variables. It's easy to show that the partition function, all marginals and all conditional MAP states of an SPN can be computed in time linear in its size. SPNs have most tractable distributions as special cases, including hierarchical mixture models, thin junction trees, and nonrecursive probabilistic context-free grammars. I will present generative and discriminative algorithms for learning SPN weights, and an algorithm for learning SPN structure. SPNs have achieved impressive results in a wide variety of domains, including object recognition, image completion, collaborative filtering, and click prediction. Our algorithms can easily learn SPNs with many layers of latent variables, making them arguably the most powerful type of deep learning to date. (Joint work with Rob Gens and Hoifung Poon.)

Pedro Domingos is Professor of Computer Science and Engineering at the University of Washington. His research interests are in machine learning, artificial intelligence and data science. He received a PhD in Information and Computer Science from the University of California at Irvine, and is the author or co-author of over 200 technical publications. He is a winner of the SIGKDD Innovation Award, the highest honor in data science. He is a AAAI Fellow, and has received a Sloan Fellowship, an NSF CAREER Award, a Fulbright Scholarship, an IBM Faculty Award, and best paper awards at several leading conferences. He is a member of the editorial board of the Machine Learning journal, co-founder of the International Machine Learning Society, and past associate editor of JAIR. He was program co-chair of KDD-2003 and SRL-2009, and has served on numerous program committees.

This talk will revolve around the innovation process we use at Netflix to improve our product, particularly the recommendation and search algorithms that help our members find movies and TV shows to watch. Our recommendation and search algorithms are key to retaining our members, so improving them has been a top priority for over a decade. Our innovation process, at its core, is an application of the scientific method, relying heavily on experimentation, very large data sets, carefully chosen metrics, and advanced statistical and mathematical models. We will discuss how our innovation process, despite its challenges and limitations, has lead to a remarkable evolution in our recommendation and search algorithms.

Carlos Gómez-Uribe is the Vice President of Product Innovation at Netflix, leading the teams responsible for the video recommendation and search algorithms that connect Netflix members with movies and TV shows to enjoy. Carlos is a scientist and engineer turned product leader with years of experience applying mathematical and statistical modeling techniques in a variety of industries. Prior to Netflix, Carlos worked at Google, McKinsey & Company, Merrill Lynch and Analog Devices. His education includes a PhD in Medical and Electrical Engineering, a Master’s Degree in Electrical Engineering and Computer Science, and Bachelor degrees in Mathematics and in Electrical Engineering and Computer Science, all from the Massachusetts Institute of Technology.

To enable computers to listen, speak, see, and learn like humans, and even more to build computers that analyze, infer, predict, and make decisions better than humans is the ultimate dream of artificial intelligence. In the near future, we envision that with advanced AI technologies each person and each enterprise will have multiple intelligent computer assistants to help accomplish various tasks with results that are as good or even better than if they were performed by humans. Huawei Technologies, one of the major telecommunication equipment and service companies in the world, is also conducting research toward this noble goal of AI. In this talk, I will introduce some progresses which Huawei, and particularly its Noah’ Ark Lab, have made recently, including a platform for deep learning research, natural language processing using deep learning, construction and utilization of large-scale knowledge base in telecommunication domain.

Hang Li is director of the Noah’s Ark Lab of Huawei Technologies. His research areas include natural language processing, information retrieval, statistical machine learning, and data mining. He graduated from Kyoto University in 1988 and earned his PhD from the University of Tokyo in 1998. He worked at the NEC lab in Japan during 1991 and 2001, and Microsoft Research Asia during 2001 and 2012.

This presentation will shed light on user tracking and behavioural targeting on the Web. Through  empirical studies of cookie tracking practices, we will take an alternative view of the display ad business by observing the network of third party trackers that envelopes the Web. The practice begs a question of how to resolve a dissonance between the current consumer tracking practices and the vendors’ desire for consumers’ loyalty and trustful long-term engagements. It also makes us aware of how computing designs and techniques, inaccessible to individuals, cause imbalance in the knowledge acquisition and enablement, disempowering the end-users.

As a Principal Researcher at Microsoft Research (MSR) in Cambridge,  Natasa is working on the design, prototyping and evaluation of information and communication systems. She is passionate about innovation in personal and social computing and promotes a dialogue between IT industry, consumers, and policy makers on the issues that arise from the adoption and use of technologies. Her current focus is on digital obsolescence, activity based computing, and privacy respecting systems and applications. Natasa is actively involved with a wider community of academics and practitioners through public speaking, collaborative projects, and serving on advisory boards of academic programs and commercial enterprises. She is a Visiting Professor at the UCL and Queen Mary University of London and a member of the ACM Europe Council. She serves on the Advisory Boards for the Course in Entrepreneurship at the University  of Cambridge and the Turing Gateway in Mathematics at the Isaac Newton Institute for Mathematical Sciences (INI).