PulseAugur
EN
LIVE 17:49:39

New Thompson Sampling methods tackle non-stationary and private contextual bandits

By PulseAugur Editorial · [3 sources] ·

Two new research papers introduce novel approaches to Thompson sampling for contextual bandits. One paper, "Flow-Corrected Thompson Sampling for Non-Stationary Contextual Bandits," proposes a Bayesian method that reuses historical data by correcting and reweighting it based on an explicit drift model, outperforming standard forgetting-based baselines. The second paper, "AdaPrivate-TS: Private Thompson Sampling for Contextual Bandits with Privacy Amplification," presents a differentially private algorithm that combines Thompson Sampling with batched zCDP composition, achieving high performance while offering strong privacy guarantees and demonstrating benefits from privacy amplification. AI

IMPACT These advancements in contextual bandit algorithms could lead to more efficient and privacy-preserving decision-making systems in areas like personalized recommendations and online advertising.

RANK_REASON Two distinct research papers published on arXiv detailing new algorithms for contextual bandits.

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 3 sources. How we write summaries →

New Thompson Sampling methods tackle non-stationary and private contextual bandits

COVERAGE [3]

  1. arXiv cs.LG TIER_1 English(EN) · AmirHossein Naghdi, Ali Baheri ·

    Flow-Corrected Thompson Sampling for Non-Stationary Contextual Bandits

    arXiv:2606.23933v1 Announce Type: cross Abstract: We study non-stationary linear contextual bandits where the reward model drifts over time, rendering classical contextual bandit algorithms brittle because historical data becomes systematically biased. We propose Flow-Corrected T…

  2. arXiv cs.LG TIER_1 English(EN) · Ali Baheri ·

    Flow-Corrected Thompson Sampling for Non-Stationary Contextual Bandits

    We study non-stationary linear contextual bandits where the reward model drifts over time, rendering classical contextual bandit algorithms brittle because historical data becomes systematically biased. We propose Flow-Corrected Thompson Sampling (fcTS), a Bayesian method that re…

  3. arXiv stat.ML TIER_1 English(EN) · Eranga Ukwatta ·

    AdaPrivate-TS: Private Thompson Sampling for Contextual Bandits with Privacy Amplification

    We present AdaPrivate-TS, a differentially private contextual bandit algorithm that combines Thompson Sampling with batched zCDP composition. Our key insight is that differential privacy noise inflates the posterior covariance in a structured way: adding Gaussian noise $N(0,σ^2 I…

RELATED ENTITIES

RELATED TOPICS