Global-Order GFlowNets

2025-04-03

AI Summary

OP-GFlowNets suffer from inter-objective conflicts in multi-objective black-box optimization due to their imposition of local Pareto ordering, undermining consistency and efficiency in Pareto front sampling. To address this, we propose Global-Order GFlowNets (GO-GFN), the first framework to lift local partial orders to a global topological order—constructing a consistent, scalarization-free global preference structure that fundamentally eliminates ordering conflicts. GO-GFN integrates probabilistic graphical modeling, global order constraints, and Pareto-front-driven flow matching for training, enabling online learning and diverse solution sampling. Evaluated on multiple multi-objective optimization (MOO) benchmarks, GO-GFN achieves significant improvements: +12.7% in Pareto coverage and +9.3% in hypervolume (HV), with enhanced convergence stability. It consistently outperforms both OP-GFlowNets and preference-conditional baselines across all metrics.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Optimization
• Constraint Satisfaction and Optimization: Constraint Optimization

Application Category

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

Abstract

Order-Preserving (OP) GFlowNets have demonstrated remarkable success in tackling complex multi-objective (MOO) black-box optimization problems using stochastic optimization techniques. Specifically, they can be trained online to efficiently sample diverse candidates near the Pareto front. A key advantage of OP GFlowNets is their ability to impose a local order on training samples based on Pareto dominance, eliminating the need for scalarization - a common requirement in other approaches like Preference-Conditional GFlowNets. However, we identify an important limitation of OP GFlowNets: imposing a local order on training samples can lead to conflicting optimization objectives. To address this issue, we introduce Global-Order GFlowNets, which transform the local order into a global one, thereby resolving these conflicts. Our experimental evaluations on various benchmarks demonstrate the efficacy and promise of our proposed method.

Problem

• Resolve conflicting objectives in Order-Preserving GFlowNets
• Transform local order into global order for optimization
• Improve multi-objective black-box optimization efficiency

Innovation

• Transforms local order into global order
• Resolves conflicting optimization objectives
• Efficiently samples diverse Pareto front candidates