Adaptive Quality-Diversity Trade-offs for Large-Scale Batch Recommendation

TL;DR

This work tackles large-scale batch recommendation by formulating a flexible quality-diversity trade-off (QDT) framework based on determinantal point processes. The authors introduce the Disentangled Quality-Diversity (DQD) family and a scalable B-DivRec (BDR) DPP that selectively filters items similar to user history via an adaptive, Nyström-based approach, enabling linear-time scaling to millions of items. An online learning layer using AdaHedge tunes the trade-off parameter λ across recommendation rounds, with a theoretical regret bound and practical demonstrations showing improved relevance and diversity under both noiseless and noisy feedback. Extensive experiments on synthetic data, MovieLens, and drug-repurposing datasets illustrate strong performance in both intrabatch and interbatch diversity, and validate the adaptive mechanism’s capacity to tailor recommendations to user responses in real time. Overall, the framework enables principled, scalable, and user-adaptive quality-diversity control for batch recommendations in large-item settings, with applicability to diverse domains including healthcare and entertainment.

Abstract

A core research question in recommender systems is to propose batches of highly relevant and diverse items, that is, items personalized to the user's preferences, but which also might get the user out of their comfort zone. This diversity might induce properties of serendipidity and novelty which might increase user engagement or revenue. However, many real-life problems arise in that case: e.g., avoiding to recommend distinct but too similar items to reduce the churn risk, and computational cost for large item libraries, up to millions of items. First, we consider the case when the user feedback model is perfectly observed and known in advance, and introduce an efficient algorithm called B-DivRec combining determinantal point processes and a fuzzy denuding procedure to adjust the degree of item diversity. This helps enforcing a quality-diversity trade-off throughout the user history. Second, we propose an approach to adaptively tailor the quality-diversity trade-off to the user, so that diversity in recommendations can be enhanced if it leads to positive feedback, and vice-versa. Finally, we illustrate the performance and versatility of B-DivRec in the two settings on synthetic and real-life data sets on movie recommendation and drug repurposing.

Figures (2)

• Figure 1: Recommendation setting in our paper.
• Figure 2: Sensitivity analysis for $\lambda$ on SYNTHETIC750 (6 users, $B=3$, $\tau=0.5$), with the MAXIMIZATION strategy.

Theorems & Definitions (4)

• Theorem 5.1
• proof
• Theorem 3.1
• proof