Authors: Haruka Kiyohara, Masahiro Nomura, Yuta Saito

Published on: February 03, 2024

Impact Score: 8.07

Arxiv code: Arxiv:2402.02171

Summary

• What is new: The introduction of the Latent IPS (LIPS) estimator for OPE in slate contextual bandits, which reduces variance without relying on linear reward functions.
• Why this is important: The high variance of the Inverse Propensity Scoring (IPS) estimator in slate contextual bandits due to large action spaces.
• What the research proposes: The Latent IPS (LIPS) estimator that optimizes slate abstractions in a low-dimensional space to minimize both bias and variance without assuming linearity in the reward function.
• Results: LIPS significantly outperforms existing estimators, especially in environments with non-linear rewards and large slate spaces.

Technical Details

Technological frameworks used: nan

Models used: Latent IPS (LIPS) estimator

Data used: nan

Potential Impact

Recommender systems, search engines, marketing, medical applications

Want to implement this idea in a business?

We have generated a startup concept here: SlateSafe Analytics.