Warm-started mixture bandits improve repeated cold-start recommendations
Challenges and Solutions for Bandits in the Wild: Warm-Started Mixture Bandits for Cross-Cohort Slate Recommendation
Machine LearningArtificial Intelligence
Summary
Many recommendation systems struggle when new users arrive without any prior history, making it hard to suggest good choices right away. The authors present a way to use data from earlier groups of users to start new recommendations smarter, so the system learns faster what each new user likes. They also design the system to avoid showing the same items too often or running out of good options. Tests in simulations and a real quiz app showed their method made better early recommendations and kept users more engaged.
What this means in practice
- •For mobile app developers: Improve early user recommendations by initializing models with data from previous user groups to tackle repeated cold-start scenarios.
- •For online game designers: Use inventory-aware recommendation slates to keep item recommendations fresh and engaging in quiz or casual game sessions with short-lived users.
Authors
Serafima Lebedeva, Sumantrak Mukherjee, Ali Arshad Sadal, Ilias Ekşi, Rahul Sharma, Julia Mueller, Theresa Dombrowski, Jakob Karolus, Viktor Bengs, Eyke Hüllermeier, Sebastian Vollmer
Abstract
Many recommender services repeatedly encounter cold-start cohorts, where new users arrive with little or no interaction history. This creates two challenges: learning user preferences quickly from limited feedback and sustaining useful recommendations when each user has a finite catalog that can become repetitive or depleted over time. We propose CohortMix-TS, a warm-started mixture bandit that learns latent user groups from earlier cohorts and uses available metadata to construct group-informed priors for new users. Starting from these fixed priors, the model personalizes independently as feedback from each user becomes available. Session slates combine Thompson sampling with diversity and inventory-depletion controls. We evaluate CohortMix-TS through simulation, semi-synthetic experiments, and a 25-day randomized in-the-wild deployment with 713 registered participants in a Campus Games quiz application. Our evaluations show that cross-cohort transfer improves early recommendation quality and user-level regret, while inventory-aware slate construction helps prevent premature exhaustion of preferred items. In the field deployment, treatment users also showed a larger early-to-late change in correctness than users receiving random recommendations. Together, these results show how warm-start transfer and inventory-aware recommendations can support personalization for short-lived, repeatedly cold-starting cohorts.