-
Won-do Lee
I am in the master's program at the Human-Centered Computing Lab, Department of Intelligent Information Convergence, Graduate School of Convergence Science and Technology, Seoul National University. Researching recommender systems, I am interested in user behavior modeling along with algorithms.
Park Min-ju
HE IS CURRENTLY ENROLLED IN THE MASTER'S PROGRAM IN THE DEPARTMENT OF INTELLIGENCE AND INFORMATION CONVERGENCE AT THE GRADUATE SCHOOL OF CONVERGENCE SCIENCE AND TECHNOLOGY, SEOUL NATIONAL UNIVERSITY, AND BELONGS TO THE MUSIC AND AUDIO LAB. DUE TO THE NATURE OF THE LAB, WE ARE VERY INTERESTED IN THE INTERFACE BETWEEN AUDIO AND AI, AND AS A PERSONAL RESEARCHER, WE ARE RESEARCHING A MUSIC RECOMMENDATION SYSTEM.
Soojin Lee
Currently, I am working as a data analyst at Woongjin ThinkBig. He mainly researches and develops recommendation systems and DKT (Deep Knowledge Tracing). He is also interested in pipelines for model serving, so he is also in charge of developing ML pipelines.
-
Those who are thinking about what a good recommendation is
People who are thinking about new methodologies every day to create a proper recommendation system
Those who are curious about the atmosphere and general contents of the RecSys 2022 conference
-
You can listen to the papers of two Korean authors accepted at RecSys 2022 in Korean.
You can gain insight into which recommendations are good recommendations and how to create a good recommendation system.
You can get insights from the actual atmosphere of the RecSys 2022 conference held in Seattle and from the perspective of practitioners.
What are the recommendations that users are enthusiastic about? There are many ways to create a recommendation system that users will be most interested in. Are users more likely to like trendy products? Is it possible to recommend a preference as a non-preference? Fortress writes a lot of any recommendations, is there anything new? There are many papers coming out, but is there anything that can actually be applied to my service? For those who are concerned, we will tell you about the experiences of the authors of RecSys 2022 and those who have been there.
agenda
20:00 ~ 20:05 Talk opening
20:05 ~ 20:25 Countering Popularity Bias by Regularizing Score Differences - Wondo Lee
The recommender system predicts a higher recommendation score for popular items that are included more in the training data, even if the items are equally liked by users. In order to reduce this popularity bias, we modified the loss function of the learning process so that the recommendation system predicts the items liked by the user with a similar recommendation score regardless of popularity. We were able to reduce the popularity bias while maintaining the accuracy of the recommendation system.
20:25 ~ 20:45 Exploiting Negative Preference in Content-based Music Recommendation with Contrastive Learning - Minjoo Park
We will share the introduction of the paper "Exploiting Negative Preference in Content-based Music Recommendation with Contrastive Learning" recently presented at RecSys 2022 and reviews of RecSys participation.
20:45 ~ 21:05 RecSys2022, the review - Soojin Lee
The RecSys 2022 conference was held in Seattle, USA in September, and I had the opportunity to attend offline. Through this time, we would like to share our overall experiences about the atmosphere of the conference, the trend of the recommendation system, and the perspectives of researchers (practitioners).
🙋🏻♀️ Are you having trouble attending the live lecture? Don't worry!
We will send a 'replay' link to those who have registered, so register now!