UMRE: A Unified Monotonic Transformation for Ranking Ensemble in Recommender Systems

2025-08-11

AI Summary

In industrial recommendation systems, conventional multi-objective ranking fusion relies on manually designed nonlinear transformations and fixed weights, resulting in high tuning costs and difficulty balancing ranking consistency with Pareto optimality. To address this, we propose the Unified Monotonic Ranking Ensemble (UMRE) framework. First, UMRE employs unconstrained monotonic neural networks (UMNNs) coupled with positive neural integration to construct a learnable, strictly monotonic score transformation module. Second, it integrates lightweight ranking models with an adaptive Pareto optimization strategy, enabling fine-grained, personalized weight fusion without manual hyperparameter tuning. Third, extensive experiments on KuaiRand and Tenrec benchmarks, along with online A/B tests, demonstrate significant improvements in NDCG, CTR, and other key metrics—validating UMRE’s effectiveness, generalizability, and production readiness.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Metareasoning and Metaheuristics
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Industrial recommender systems commonly rely on ensemble sorting (ES) to combine predictions from multiple behavioral objectives. Traditionally, this process depends on manually designed nonlinear transformations (e.g., polynomial or exponential functions) and hand-tuned fusion weights to balance competing goals -- an approach that is labor-intensive and frequently suboptimal in achieving Pareto efficiency. In this paper, we propose a novel Unified Monotonic Ranking Ensemble (UMRE) framework to address the limitations of traditional methods in ensemble sorting. UMRE replaces handcrafted transformations with Unconstrained Monotonic Neural Networks (UMNN), which learn expressive, strictly monotonic functions through the integration of positive neural integrals. Subsequently, a lightweight ranking model is employed to fuse the prediction scores, assigning personalized weights to each prediction objective. To balance competing goals, we further introduce a Pareto optimality strategy that adaptively coordinates task weights during training. UMRE eliminates manual tuning, maintains ranking consistency, and achieves fine-grained personalization. Experimental results on two public recommendation datasets (Kuairand and Tenrec) and online A/B tests demonstrate impressive performance and generalization capabilities.

Problem

• Replacing manual nonlinear transformations in ensemble sorting
• Balancing competing objectives without labor-intensive tuning
• Achieving Pareto efficiency and personalized ranking in recommendations

Innovation

• UMNN replaces manual transformations with neural integrals
• Lightweight ranking model for personalized score fusion
• Pareto optimality strategy adaptively balances task weights