Generative approach guides design toward optimal multi-goal tradeoffs
ParetoTransport: Generative Optimization by Mass Transport Toward The Pareto Front
Machine Learning
Summary
When designing things that need to be good in many different ways at once, it’s hard to find choices that balance all goals well. The authors created a way for computer models to move groups of design options closer to the best set of tradeoffs, called the Pareto front, by shaping their overall distribution rather than working on each option separately. Their method, ParetoTransport, uses a mathematical approach to gently push design options closer to balance in a smooth and controlled way, improving how well designs meet multiple goals at once. Testing shows this method works better than previous ones on several common benchmarks.
What this means in practice
- •For engineering design teams: Generate balanced design options that simultaneously optimize multiple goals with better coverage and quality along the tradeoff front.
- •For automotive development teams: Improve offline optimization for vehicle components by efficiently steering candidate designs towards the best multi-criteria compromises.
Authors
Stephanie Holly, Sepp Hochreiter, Werner Zellinger
Abstract
Offline multi-objective optimization requires not only moving the objective vectors of candidate designs toward the Pareto front, but also distributing them effectively along it. Generative methods have recently emerged as a natural approach because they learn a distribution over feasible designs while allowing generation to be steered toward promising designs. Existing methods, however, largely retain classical sample-wise guidance strategies, leaving the distribution-level modeling capability of generative methods underused. We propose ParetoTransport, a training-free guidance method for pre-trained flow-matching models that explicitly specifies and refines a population-level distribution in objective space. ParetoTransport guides a flow-matching sampler to iteratively transport the empirical offline distribution toward the Pareto front, with Wasserstein matching to intermediate proxy distributions. This directly controls distributional displacement and mass allocation along the front. We establish a convergence result and demonstrate state-of-the-art performance on standard offline MOO benchmarks, extending recent evaluations beyond hypervolume to generational distance, inverted generational distance, and Wasserstein distance.