"Reinforcement Learning for Recommender Systems: A Case Study on Youtube," by Minmin Chen
Vložit
- čas přidán 9. 07. 2024
- While reinforcement learning (RL) has achieved impressive advances in games and robotics, it has not been widely adopted in recommender systems. Framing recommendation as an RL problem offers new perspectives, but also faces significant challenges in practice. Industrial recommender systems deal with extremely large action spaces - many millions of items to recommend and complex user state spaces -- billions of users, who are unique at any point in time. In this talk, I will discuss our work on scaling up a policy-gradient-based algorithm, i.e. REINFORCE to a production recommender system at CZcams. We proposed algorithms to address data biases when deriving policy updates from logged implicit feedback. I will also discuss some follow up work and outstanding research questions in applying RL, in particular off-policy optimization in recommender systems.
- Věda a technologie
I love this field of research. I have had to rely on youtube almost entirely to compensate for the basement quality education at my university. I love the possibilities of this, I just wish I had the fundamental grasp of the granular mechanics to go into researching it myself.
Thanks for sharing this interesting topic in real cases. Many thanks
Any link to the related open-source implementation?
Is there any video about the talks in Wednesday 16:30 she mentioned?
i haven't found one. It was a 15 minute presentation on a paper called "Top-K Off-Policy Correction for a REINFORCE Recommender System." if you're interested. there seems to be a video about that
Great talk. Is the Wednesday talk also somewhere in youtube?
thank you , i am interest recomemder system for finish my project
Please, wednesday talk requested.
厉害,大盘涨了0.86,接近1个点。
Very interesting, although I followed at 1.25x speed because the speaking pace was a bit slow. I'll definitely check out the mentioned Wednesday talk!
A small suggestion: the questions were hardly audible, writing them on screen would help
It's open mind to me to use RL for recommender system. And Minmin is cute. : )
Cute!