AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

RaCT: Toward Amortized Ranking-Critical Training For Collaborative Filtering

Sam Lobel*, Chunyuan Li*, Jianfeng Gao, Lawrence Carin

ICLR 2020 · Reinforcement learning › Policy gradient and actor-critic · approach: Actor-critic theory · 9 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Apply RaCT to VAEVAE · ML-20MNDCG@100: 0.4258 -> 0.4339Improved
Apply RaCT to VAEVAE · NetflixNDCG@100: 0.385 -> 0.392Improved
Apply RaCT to VAEVAE · MSDNDCG@100: 0.310 -> 0.319Improved
Replace the multinomial likelihood with a Gaussian likelihood (ablation)VAE actor + RaCT · ML-20MNDCG@100: 0.4339 -> 0.4224Hurt
Remove KL regularization and use a deterministic actor (ablation)VAE actor + RaCT · ML-20MNDCG@100: 0.4339 -> 0.4255Hurt
Use a linear encoder and decoder (ablation)VAE actor + RaCT · ML-20MNDCG@100: 0.4339 -> 0.4162Hurt
Remove the held-out interaction count feature |H0| (ablation)Feature-based critic · ML-20MNDCG@100Hurt
Remove the observed interaction count feature |H1| (ablation)Feature-based critic · ML-20MNDCG@100Hurt
Remove both interaction-count features (ablation)Feature-based critic · ML-20MNDCG@100Hurt
Train the critic with Recall@100 instead of NDCG@100 (ablation)RaCT · ML-20MRecall@20: 0.40269 -> 0.40316Improved
Train the critic with Recall@100 instead of NDCG@100 (ablation)RaCT · ML-20MNDCG@100: 0.43395 -> 0.43392No clear effect
Apply RaCT to WARPWARP actor · ML-20MNDCG@100: 0.3123 -> 0.3439Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy