# Agenda and Meeting 02/2/2022 ## Recsim We have gone through the tutorials to get a general understanding. The lack of documentation is a problem. We talked about implementing our own rl-rs to get a better understanding, using the movielens database. We looked at LensKit, but it seems to be a general framework for RS but not for RL. - Dopamine / other frameworks for simulating RLRS? - what's worth using? - example programs or documents or simulation of recsim - If we find something, is it better to dive in deep or is it better to explore? ## Reservations - Basically: lack of motivation due to the vagueness of the whole project. - Even when we are motivated it's hard to know what to do? -> any solutions/ recommendations for this? - Even now we don't feel like we are at a level where we could come up with a research hypothesis, At least on my part, I'm concerned with the lack of results. Feels like we get short motivation bursts which quickly fade away. And even when we do have time to work on it, I do not know what to do. I guess recsim could resolve that. Lack of goals. - What's up with PhD's? We think that would help resolve these problems. - How to pick papers to dive into? - Lack of progress ## Meeting Toy around with recsim. Find exmaples online for gym Getting busy running there, do something feasible 1. Start with setting up a system for recsim Research topics: - Dealing with lots of actions that are very similar. Cluster actions? https://arxiv.org/abs/1905.12767 Next meeting: two weeks from now There's a master student. lmao