MLRec 2018

MLRec 2018

4^th International Workshop on Machine Learning Methods for Recommender Systems

In conjunction with 18th SIAM International Conference on Data Mining (SDM 2018)
May 3 - 5, 2018, San Diego, CA, USA

CALL FOR PAPER

Following the success of the several editions of MLRec in 2015, 2016, 2017, the fourth edition of the MLRec workshop focuses on developing novel, and applying existing Machine Learning (ML) and Data Mining (DM) methods to improve recommender systems. This workshop also highly encourages applying ML-based recommendation algorithms in novel application domains (e.g., precision medicine), deep learning for recommendation, and solving novel recommendation problems formulated from industry. The ultimate goal of the MLRec workshop series is to promote the advancement and implementation of new, effective and efficient ML and DM techniques with high translational potential for real and large-scale recommender systems, and to expand the territory of ML-based recommender system research toward non-conventional application areas where recommendation problems largely exist but haven't been fully recognized.

Topics of Interest

We encourage submissions on a variety of topics, including but not limited to:

• Novel machine learning algorithms for recommender systems, e.g., new content-based or context-aware recommendation algorithms, new algorithms for matrix factorization, tensor-based approaches for recommender systems, etc.
• Novel applications of existing machine learning and data mining algorithms for recommender systems, e.g., applying bilinear models, (non-convex) sparse learning, metric learning, low-rank approximation/PCA/SVD, neural networks and deep learning, etc.
• Novel optimization techniques for improving recommender systems, e.g., parallel/distributed optimization techniques, efficient stochastic gradient descent, etc.
• Industrial practices and implementations of recommendation systems, e.g., feature engineering, model ensemble, large-scale implementations of recommender systems, etc.
• Emerging recommendation problems and scenarios in industry and their ML-based solutions, e.g., recommendation for e-fashion, etc.
• Novel recommendation problems in non-conventional recommender system research areas (e.g., precision medicine, health informatics) and their ML-based solutions, e.g., recommendation of physicians, recommendation of healthy life-styles for seniors, etc.
• Enhanced deep learning methods for recommender systems, e.g., word embedding techniques, CNN, RNN and LSTM, Generative Advertiseral Networks (GAN), auto-encoder, RBM, etc.
• Recommendation in Information Retrieval, ad industry, targeting ad, search ad, etc.

Submission

Instructions

The workshop accepts long paper and short (demo/poster) papers. Short papers submitted to this workshop should be limited to 4 pages while long papers should be limited to 8 pages. All papers should be formatted using the SIAM SODA macro. Authors are required to submit their papers electronically in PDF format to the submission site by 11:59pm MDT, Jan 15, 2018. The site has started to accept manuscripts. At least one author of each accepted paper should be registered to the conference.

Important Dates

• Paper Submission Deadline: December 23 Jan 15, 2018
• Author Notification: January 23, 2018
• Camera Ready Paper Due: February 1, 2018
• Workshop: May 5, 2018

Invited Speakers

Martin Ester, Professor, Simon Fraser University
Title: Recommendation in Social Media
Abstract: Social media are media for social interaction, using highly accessible and scalable communication techniques to create and exchange user-generated content. In the social sciences, the effects of social influence, homophily or selection, and transitivity have been identified as drivers of the dynamics of social networks. In the first part of the talk we will present a matrix factorization approach to incorporate these effects into a recommender system by regularizing the user factors with the user factors of direct friends. In the second part, we will discuss how to exploit product reviews, a popular type of user-generated content, to improve the accuracy of recommendations. Our approach combines probabilistic matrix factorization with aspect-based opinion mining, and we propose a graphical model based on latent dirichlet allocation (LDA) that extracts product aspects and their ratings. Another important aspect of social media is their support for user groups. In the third part of the talk, we explore the recommendation of user groups. We propose measures for the different types of engagement of users with groups that are more suitable than simple group membership. We also introduce a temporal matrix factorization method that models the dynamically changing patterns of user group engagement. The talk will conclude with the discussion of interesting directions for future research.
Bio: Martin Ester received a PhD in Computer Science from ETH Zurich, Switzerland, in 1990 with a thesis on knowledge-based systems and logic programming. He has been working for Swissair developing expert systems before he joined University of Munich as an Assistant Professor in 1993. Since November 2001, he has been an Associate Professor, now Full Professor at the School of Computing Science of Simon Fraser University, where he co-directs the Database and Data mining research lab. He has published extensively in the top conferences and journals of his field such as KDD, ICDM and TKDE, and his work has been very well-cited. His most famous paper on DBSCAN received more than 6600 citations, and his H-index is 47. He recently served as PC Co-Chair of SDM 2018, ACM/IEEE ASONAM 2014 and ACM RecSys 2014. His current research interests include social network analysis, recommender systems, opinion mining, biological network analysis and high-throughput sequence data analysis.

Julian McAuley, Assistant Professor, University of California, San Diego
Title: Structured Output Models of Recommendations, Activities, and Behavior
Abstract: Predictive models of human behavior--and in particular recommender systems--learn patterns from large volumes of historical activity data, in order to make personalized predictions that adapt to the needs, nuances, and preferences of individuals. Models may take incredibly complex data as *input*, ranging from text, images, social networks, or sequence data. However, the *outputs* they are trained to predict--clicks, purchases, transactions, etc.--are typically simple, numerical quantities, in order for the problem to be cast in terms of traditional supervised learning frameworks. In this talk, we discuss possible extensions to such personalized, predictive models of human behavior so that they are capable of predicting complex structured *outputs*. For example, rather than training a model to predict what content a user might interact with, we could predict how they would react to unseen content, in the form of text they might write. Or, rather than predicting whether a user would purchase an existing product, we could predict the characteristics or attributes of the types of products that *should* be created.
Bio: Julian McAuley’s research focuses on the linguistic, temporal, and visual dimensions of opinions and behavior in social networks and online communities. This includes understanding the facets of people's opinions, the processes by which people "acquire tastes" for gourmet foods and beers, or even the visual dimensions that make clothing items compatible. He has been an assistant professor at UC San Diego since 2014, and received his PhD from the Australian National University.

Neel Sundaresan, Partner Director, Microsoft
Title: Practical Recommender Systems: It's Not Just about Algorithms
Abstract: I will talk about approaches and experiences in building, deploying, testing and scaling recommender systems in commercial contexts. Recommender systems are used extensively in online shopping to campaign management to desktop software systems. While it is important to build lever data driven algorithms , several aspects of scale, cost, speed, product and revenue goals require one to look at other aspects of deploying such systems. We will discuss real-life examples of such industrial deployments.
Bio: Dr. Neel Sundaresan is now a partner director at Microsoft. He is an entrepreneur and advanced technologist with extensive experience in both the startup world and large companies. He has rich executive level engineering, product, and research experience managing varied engineering, and research organizations and building organizations from the ground up. He has significant research and technology experience with 70+ publications and 100 issued patents and several pending. He has extensive experience starting up and growing strong organizations that are world-class innovative, conduct deep science, and execute at the highest level. He built out a worldclass economics team at eBay; founded the most creative computer vision team; a topnotch big data science and machine learning teams.

Suju Rajan, VP, Head of Research at Criteo
Title: Modeling user intent for recommendation
Abstract: In this talk we will go over two case studies, one from personalization of the newsfeed and another on product recommendation in a demand side platform. In both cases, the intention is to provide personalized recommendations to the users but the path to designing such systems is often not straight-forward and a lot of assumptions are made on the way. Through the two applications, I hope to draw out and motivate challenges that one faces in real-world large-scale recommendation and highlight some attempts at solving these problems.
Bio: Suju Rajan is the VP, Head of Research at Criteo. At Criteo, her team works on all aspects of performance driven computational advertising, including, real-time bidding, large-scale recommendation systems, auction theory, reinforcement learning, online experimentation, metrics and scalable optimization methods. Prior to Criteo, she was the Director of the Personalization Sciences at Yahoo Research where her team worked on personalized recommendations for several Yahoo products. She received her PhD from the University of Texas at Austin, focusing on semi-supervised and active learning based classification for dynamic environments.

Huzefa Rangwala, Associate Professor, George Mason University
Title: Recommender Systems in Educational Data Mining
Abstract: The application of big data approaches, specifically methods inspired from recommender system domain to predict student performance is largely a new area of research. The types of solutions available depend largely on the type of available data, and problem definition. For instance, for the purposes of degree planning, one task is to predict grades for a student in a class in the future (or in the next term). In this talk, I will present an overview of problems, challenges and solutions for solving critical problems faced by higher education institutions.
Bio: Huzefa Rangwala is an Associate Professor at the Department of Computer Science & Engineering, George Mason University. He was a Visiting Faculty Member at Department of Computer Science, Virginia Tech in 2015-2016. His research interests include data mining, learning analytics, recommend systems, bioinformatics and high performance computing. He is the recipient of the NSF Early Faculty Career Award in 2013, the 2014 GMU Teaching Excellence Award, the 2014 Mason Emerging Researcher Creator and Scholar Award, the 2013 Volgenau Outstanding Teaching Faculty Award, 2012 Computer Science Department Outstanding Teaching Faculty Award and 2011 Computer Science Department Outstanding Junior Researcher Award. His research is funded by NSF, NIH, NRL, DARPA, USDA and nVidia Corporation.

Dawen Liang, Research Scientist, NetFlix
Title: Beyond linear latent factor models for collaborative filtering
Abstract: Linear latent factor models still largely dominate the collaborative filtering research literature due to their simplicity and effectiveness. Exploring more powerful non-linear counterpart is an active research area. In this talk, I will outline two related projects: First, I will show that by incorporating carefully designed non-linear features into the classical matrix factorization model, we are able to significantly boost the performance (RecSys'16); Second, I will talk about our recent work on extending variational autoencoder, a non-linear probabilistic model, to collaborative filtering with state-of-the-art results (WWW'18).
Bio: Dr. Liang is a research scientist at Netflix. He completed my Ph.D. in the Electrical Engineering Department at Columbia University, as part of the LabROSA, working with Professor Dan Ellis and Professor David Blei.

Organizers

Organizing Committee

Deguang Kong Email is currently a Sr. Research Scientist at Yahoo Research, leading industry-level ad science and recommendation project. He has strong background in machine learning/data mining, and published over 30 referred articles in acadmia conferences (such as ICML, NIPS, AAAI, CVPR, KDD, ICDM, SDM, WSDM, CIKM, etc) and also led mobile data science efforts at Samsung during 2014-2015. He served as a PC member for top conferences and journals, such as KDD, SDM, NIPS, TKDD, etc. Deguang Kong has been on the organization panel for MLRec2015, MLRec 2016, and MLRec 2017.

Xia Ning Email is currently an Assistant Professor at the Department of Computer and Information Science, Indiana University – Purdue University Indianapolis (IUPUI). Before joining IUPUI, she was a research staff member at NEC Labs America. Xia received her PhD. degree in Computer Science at University of Minnesota, Twin Cities in 2012. Her research focuses on Recommender Systems, Chemical Informatics and Health Informatics. The results from her research have been presented in various conferences and published in leading peer reviewed journals and highly selective conference proceedings. She has been serving as a program committee member on various premier data mining conferences such as KDD, ICDM and SDM, and Recsys.

George Karypis Email is currently Professor at the Department of Computer Science & Engineering at the University of Minnesota in the Twin Cities of Minneapolis and Saint Paul and a member of the Digital Technology Center (DTC) at the University of Minnesota. His research interests are concentrated in the areas of bioinformatics, cheminformatics, data mining, and high-performance computing, and from time-to-time, he looks at various problems in the areas of information retrieval, collaborative filtering, and electronic design automation for VLSI CAD.Within these areas, his research focuses in developing novel algorithms for solving important existing and/or emerging problems, and on developing practical software tools implementing some of these algorithms. The results from his research have been presented in various conferences and published in leading peer reviewed journals and highly selective conference proceedings

Program

Saturday, May 5th

Accepted Papers