The Machine Learning & Inference Research team I co-lead @Netflix @NetflixResearch is hiring interns for Summer 2024. Looking for a research internship (tackling industry problems while also focusing publishable research!)? Apply thru this listing: jobs.netflix.com/jobs/300628646

Remember “LGBT free” zones in Poland? New research finds that annual suicide attempts increased by 16%, or 5 attempts per 100k capita after these laws were enacted! www.nber.org/papers/w31702

arxiv.org/abs/2302.02392 In offline RL, we replace exploration with assumptions that data is nice. We try to make these minimal by refining standard realizability and coverage assumptions to single policies. We do this via a minimax formulation and strong guarantees for learning the saddle point.

arxiv.org/abs/2305.15703 RL only needs mean reward to go (q-fn) so why is distRL (learn whole reward-to-go dist) so empirically effective? We prove distRL is really good when optimal policy has small loss. When that's true then least-squares (q-learning) misses the signal due to heteroskedasticity.

arxiv.org/abs/2207.13081 Off-policy eval in POMDPs is tough b/c hidden states ruin memorylessness inducing a curse of horizon. Using histories as instrumental variables, we derive a new Bellman eq for a new kind of v-fn. We solve it using minimax learning to get model-free eval using general fn apx.

New platform just in time for shameless plugs of neurips papers! (Tho hoping to use this platform for more than shameless paper plugs... let's see.) Genuinely excited about each of these. Let me try to get you excited about them too with a couple sentences each.