KDD OARS Workshop

Introduction

The KDD workshop on online and adaptive recommender systems (OARS) will serve as a platform for publication and discussion of OARS. This workshop will bring together practitioners and researchers from academia and industry to discuss the challenges and approaches to implement OARS algorithms and systems, and improve user experiences by better modeling and responding to user intent.

Many recommender systems deployed in the real world rely on categorical user-profiles and/or pre-calculated recommendation actions that stay static during a user session. Recent trends suggest that recommender systems should model user intent in real time and constantly adapt to meet user needs at the moment or change user behavior in situ. In addition, various techniques have been proposed to help recommender systems adapt to new users, items, or behaviors. Some strategies to build “adaptive” recommenders include:

Systems for online training, e.g. updating the parameters of a pre-trained model to a new user.
Feature-based systems that handle cold-start scenarios, and can gracefully adapt to a combination of cold- and warm users/items.
Systems that avoid modeling users at all (e.g. session-based recommenders that directly learn from item interactions without needing user terms)
Systems that adapt to new behaviors through RL or other adaptive learning algorithms.

We invite submission of papers and posters of two to ten pages (including references), representing original research, preliminary research results, proposals for new work, position, and opinion papers. All submitted papers and posters will be single-blind and will be peer reviewed by an international program committee of researchers of high repute. Accepted submissions will be presented at the workshop.

Topics of interest include, but are not limited to:

Novel algorithms and paradigms
- online and adaptive neural recommender
- reinforcement learning (on-policy, off-policy, offline RL, and other relevant subfields)
- online/streaming learning
- interactive and conversational recommender
- extreme classification
- graph recommender
Applications
- product recommendations
- content recommendations (news, music, movie, video, etc.)
- ads recommendations
- fashion and decor recommendation
- job recommendation
- intervention/behavior change/healthy life-style recommendations
User modeling and representations
- implicit and explicit user intent modeling
- dynamic user intent modeling
- visual/style/taste modeling
- combination of in-session intent with long term user interest
- incorporation of knowledge graph
- representation learning
Architecture and infrastructure
- scalability of neural methods for large scale real-time recommendations
- steaming and event-driven processing infrastructures
Evaluation and explanation methodologies
- evaluation, comparison, explanation of OARS for a recommendation task
- off-policy and counterfactual evaluation
Social and user impact
- UX for OARS
- welfare and objectives of OARS (CTR, dwell-time, diversity, multi-objectives, long term objectives)
- privacy and ethics considerations

Call for Papers

Download CFP

All papers will be peer reviewed (single-blind) by the program committee and judged by their relevance to the workshop, especially to the main themes identified above, and their potential to generate discussion.

All submissions must be formatted according to the latest ACM SIG proceedings template (two column format). One recommended setting for Latex file of manuscript is: \documentclass[sigconf, anonymous, review]{acmart}. Submissions must describe work that is not previously published, not accepted for publication elsewhere, and not currently under review elsewhere. All submissions must be in English.

Please note that at least one of the authors of each accepted paper must register for the workshop and attend the online session to present the paper during the workshop.

Submissions to KDD OARS workshop should be made at easychair page.

Key Dates

Submissions Due	~~May 26~~June 5
Notification	June 20
Camera Ready Version Due	Aug 1
Workshop Day	Aug 14

Feedback and stay updated to the workshop

Schedule

		August 14
8:00	-8:10 ET	Openning
8:10	-8:50 ET	Invited Talk 1 Secure and Private Recommendation Tommaso Di Noia, Polytechnic University of Bari Security and privacy of recommender systems are the two sides of the same coin representing the user. Attacking a recommendation engine to provide malicious suggestions or get users’ data may result in a bad user experience. In this talk, we will touch on the techniques and approaches to attack and defend a recommender system and preserve the users' privacy while keeping high accuracy. In particular, we will focus on the “security side of the coin” and provide a few hints on the techniques and approaches adopted in the literature to cope with privacy in recommendation scenarios.
8:50	-9:30 ET	Invited Talk 2 Learning from a learning user: a new perspective for interactive recommendation Prof. Hongning Wang, University of Virginia Existing research of recommender systems relies on an overly simplified user model: users are omniscient about the item space. However, such a static and explicit user behavior assumption appears less realistic in many modern recommendation scenarios, where the prohibitively large candidate item space prevents any individual to get the complete knowledge ahead of time. In our recent work, we propose a refined problem setting to study the sequential interactions between a recommender system and its users. We focus on the users’ learning behaviors, where a user also has to update her utility estimation based on rewards collected from the accepted recommendations to make improved choices. We study the problem of no-regret learning on the system side. And our findings suggest the intrinsic difficulty introduced by the user’s learning behavior and the possibility of efficient online learning algorithm design for the system. This provides new insights on practical recommender system design.
9:30	-10:10 ET	Invited Talk 3 TorchRec: a library for modern production recommendation systems Colin Taylor, Meta AI Recommendation Systems (RecSys) comprise a large footprint of production-deployed AI today. For researcher and ML practitioners alike, the field is far from democratized. Further, RecSys as an area is largely defined by learning models over sparse and/or sequential events, which has large overlaps with other areas of AI. Many of the techniques are transferable, particularly for scaling and distributed execution. A large portion of the global investment in AI is in developing these RecSys techniques, so cordoning them off blocks this investment from flowing into the broader AI field. In this talk we introduce TorchRec, a PyTorch domain library for Recommendation Systems. This new library provides efficient distributed embedding table primitives, enabling researchers to build state-of-the-art personalization models and deploy them in production. TorchRec library is currently used to train most large-scale recommender models at Meta including Ads, and Facebook and Instagram Reels. In this talk we cover how Torchrec enables scaling, optimizations for embedding lookup and compute, embedding layer optimizer fusion and production deployment.
10:10	-10:50 ET	Invited Talk 4 Building and Understanding Recommenders for Long-Term User Experiences Alex Beutel, Google How do our ML design choices align with real applications and shape user experiences? In recent years, our understanding of the objective of recommenders has evolved from making individual good predictions of user interest, to now creating positive, long-term experiences. This task of enabling positive, long-term experiences is significantly more challenging, both making clear the recommender responsibility in shaping the user experience and the difficulty in optimizing over long-term user trajectories. In this talk we'll discuss the ways in which taking this longitudinal view effects how we formulate our task, design our machine learning, and evaluate our systems. We'll discuss recent advances and challenges in adapting reinforcement learning for recommendation, and designing simulations to better understand recommender systems.
10:50	-11:05 ET	Contributed Talk 1 Dynamic Memory for Interpretable Sequential Optimisation Srivas Chennu, Andrew Maher, Jamie Martin and Subash Prabanantham
11:05	-11:20 ET	Contributed Talk 2 Challenges for Session-based Recommender Systems in next generation IFE-Systems Marko Harasic and Adrian Paschke
11:20	-11:35 ET	Contributed Talk 3 A Performance-preserving Fairness Intervention for Adaptive Microfinance Recommendation Robin Burke, Pradeep Ragothaman, Nicholas Mattei, Brian Kimmig, Amy Voida, Nasim Sonboli, Anushka Kathait and Melissa Fabros
11:35	-11:50 ET	Contributed Talk 4 Bootstrapping Interactive Recommender Systems Sally Goldman, Dima Kuzmin, Steffen Rendle, Li Zhang, Moustafa Alzantot, Ajit Apte, Ajay Joshi, Anand Kesari, Santiago Ontanon, Anushya Subbiah, Martin Zinkevich, John Anderson and Sarvjeet Singh
11:50	-12:30 ET	Posters Photos Are All You Need for Reciprocal Recommendation in Online Dating James Neve and Ryan Mcconville Weighing Dynamic Availability and Consumption for Twitch Recommendations Edgar Chen, Saad Ali, Eder Santana and Mark Ally Adaptive real-time diversification of digital content Tushar Shandhilya and Nisheeth Srivastava Beyond Standard Performance Measures in Extreme Multi-label Classification Erik Schultheis, Marek Wydmuch, Rohit Babbar and Krzysztof Dembczyński