Prof. Inderjit S. DhillonGottesman Family Centennial Professor, Director, Center for Big Data Analytics, Department of Computer Science, University of Texas at Austin
Multi-Output Prediction: Theory and Practice
Many challenging problems in modern applications amount to finding relevant results from an enormous output space of potential candidates, for example, finding the best matching product from a large catalog or suggesting related search phrases on a search engine. The size of the output space for these problems can be in the millions to billions. Moreover, observational or training data is often limited for many of the so-called “long-tail” of items in the output space. Given the inherent paucity of training data for most of the items in the output space, developing machine learned models that perform well for spaces of this size is challenging. Fortunately, items in the output space are often correlated thereby presenting an opportunity to alleviate the data sparsity issue. In this talk, I will first discuss the challenges in modern multi-output prediction, including missing values, features associated with outputs, absence of negative examples, and the need to scale up to enormous data sets. Bilinear methods, such as Inductive Matrix Completion~(IMC), enable us to handle missing values and output features in practice, while coming with theoretical guarantees. Nonlinear methods such as nonlinear IMC and DSSM (Deep Semantic Similarity Model) enable more powerful models that are used in practice in real-life applications. However, inference in these models scales linearly with the size of the output space. In order to scale up, we present the Prediction for Enormous and Correlated Output Spaces (PECOS) framework, that performs prediction in three phases: (i) in the first phase, the output space is organized using a semantic indexing scheme, (ii) in the second phase, the indexing is used to narrow down the output space by orders of magnitude using a machine learned matching scheme, and (iii) in the third phase, the matched items are ranked by a final ranking scheme. The versatility and modularity of PECOS allows for easy plug-and-play of various choices for the indexing, matching, and ranking phases, and it is possible to ensemble various models, each arising from a particular choice for the three phases.
Inderjit Dhillon is the Gottesman Family Centennial Professor of Computer Science and Mathematics at UT Austin, where he is also the Director of the ICES Center for Big Data Analytics. Currently he is on leave from UT Austin and heads the Amazon Research Lab in Berkeley, California, where he is developing and deploying state-of-the-art machine learning methods for Amazon Search. His main research interests are in big data, deep learning, machine learning, network analysis, linear algebra and optimization. He received his B.Tech. degree from IIT Bombay, and Ph.D. from UC Berkeley. Inderjit has received several awards, including the ICES Distinguished Research Award, the SIAM Outstanding Paper Prize, the Moncrief Grand Challenge Award, the SIAM Linear Algebra Prize, the University Research Excellence Award, and the NSF Career Award. He has published over 200 journal and conference papers, and has served on the Editorial Board of the Journal of Machine Learning Research, the IEEE Transactions of Pattern Analysis and Machine Intelligence, Foundations and Trends in Machine Learning and the SIAM Journal for Matrix Analysis and Applications. Inderjit is an ACM Fellow, an IEEE Fellow, a SIAM Fellow and an AAAS Fellow.
Prof. Samuel KaskiProfessor, Computer Science, Aalto University
Data analysis with humans ( Download Slides )
Data analysis is usually done by humans or for the use of humans. Why, then, do we not include humans in the models we use for data analysis? I will suggest three ways to improve modelling results by taking the human user into account in probabilistic data analysis, by joint modelling of the user and the domain data. First, the user can be a source of prior knowledge, resulting in interactive prior elicitation. Second, the user can be a special type of data source, resulting in extending active learning. Third, the user can be the final decision maker, who requires understandable and relevant results in the AI-assisted modelling and design process. Further complications for modelling arise from users being active planners, instead of passive data sources as often thought, and having their limitations.
Samuel Kaski is an academy (research) professor and professor of computer science at Aalto University. He leads the Finnish Center for Artificial Intelligence FCAI and the starting ELLIS Unit Helsinki. His field is probabilistic machine learning, with applications involving multiple data sources in interactive information retrieval, user interaction, health and biology. He is an action editor of two top machine learning journals (JMLR, IEEE TPAMI), and has chaired several conferences including AISTATS 2014 and ACM IUI 2022. He has published 285 peer-reviewed papers, of which several have won best-paper awards, and supervised 24 PhD theses. Of his lab alumni, 17 have got faculty positions.
Prof. Jure LeskovecAssociate Professor, Computer Science at Stanford University
Advancements in Graph Neural Networks ( Download Slides )
Machine learning on graphs is an important and ubiquitous task with applications ranging from drug design to friendship recommendation in social networks. The primary challenge in this domain is finding a way to represent, or encode, graph structure so that it can be easily exploited by machine learning models. In this talk I will discuss recent advancements in the field of Graph Neural Networks that automatically learn to encode graph structure into low-dimensional embeddings, using techniques based on deep learning. I will provide a conceptual overview of key advancements in this area of representation learning on graphs, including graph convolutional networks and their representational power. We will also discuss applications to web-scale recommender systems, healthcare, and knowledge representation and reasoning.
Jure Leskovec is Associate Professor of Computer Science at Stanford University, Chief Scientist at Pinterest, and investigator at Chan Zuckerberg Biohub. His research focuses on machine learning and data mining with graphs, a general language for describing social, technological and biological systems. Computation over massive data is at the heart of his research and has applications in computer science, social sciences, marketing, and biomedicine. This research has won several awards including a Lagrange Prize, Microsoft Research Faculty Fellowship, the Alfred P. Sloan Fellowship, and numerous best paper and test of time awards. Leskovec received his bachelor's degree in computer science from University of Ljubljana, Slovenia, PhD in machine learning from Carnegie Mellon University and postdoctoral training at Cornell University.
Prof. Bing LiuDistinguished Professor, Department of Computer Science, University of Illinois at Chicago (UIC)
Open-World AI and Continual Learning
The classic machine learning (ML) paradigm learns in isolation, which is only suitable for well-defined narrow tasks in closed environments. It is far from sufficient for AI systems such as self-driving cars and chatbots that need to work in the real-world dynamic and open environment that is full of unknowns. We humans learn comfortably in such environments. We learn continuously, accumulate the learned knowledge, and learn more and better in a self-motivated and self-supervised manner in our interactions with other humans and the physical environment. Lifelong/continual learning in the open world aims to imitate this human continuous learning capability. In this talk, I will discuss open-world AI and continual learning using self-driving cars and chatbots as applications. These systems seem to need the human-level of intelligence to work well, which the current AI and/or ML algorithms are unable to provide.
Bing Liu is a distinguished professor of Computer Science at the University of Illinois at Chicago (UIC). He received his Ph.D. in Artificial Intelligence (AI) from the University of Edinburgh. Before joining UIC, he was a faculty member at the School of Computing, National University of Singapore (NUS). His research interests include lifelong and continual learning, sentiment analysis, chatbots, open-world AI/learning, natural language processing (NLP), and data mining and machine learning. He has published extensively in top conferences and journals. He also authored four books: two on sentiment analysis, one on lifelong learning, and one on Web mining. Three of his papers have received Test-of-Time awards: two from SIGKDD (ACM Special Interest Group on Knowledge Discovery and Data Mining), and one from WSDM (ACM International Conference on Web Search and Data Mining). Some of his work has also been widely reported in the international press, including a front-page article in the New York Times. On professional services, he has served as the Chair of ACM SIGKDD from 2013-2017, as program chair of many leading data mining conferences, including KDD, ICDM, CIKM, WSDM, SDM, and PAKDD, as associate editor of leading journals such as TKDE, TWEB, DMKD and TKDD, and as area chair or senior PC member of numerous NLP, AI, Web, and data mining conferences. He is a recipient of ACM SIGKDD Innovation Award (the most prestigious technical award from SIGKDD), and he is also a Fellow of the ACM, AAAI, and IEEE.