SPREAD: Sampling-based Pareto front Refinement via Efficient Adaptive Diffusion

2025/09/25 by Hotegni, Sedjro Salomon, Peitz, Sebastian
#FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2509.21058

Abstract

Developing efficient multi-objective optimization methods to compute the Pareto set of optimal compromises between conflicting objectives remains a key challenge, especially for large-scale and expensive problems. To bridge this gap, we introduce SPREAD, a generative framework based on Denoising Diffusion Probabilistic Models (DDPMs). SPREAD first learns a conditional diffusion process over points sampled from the decision space and then, at each reverse diffusion step, refines candidates via a sampling scheme that uses an adaptive multiple gradient descent-inspired update for fast convergence alongside a Gaussian RBF-based repulsion term for diversity. Empirical results on multi-objective optimization benchmarks, including offline and Bayesian surrogate-based settings, show that SPREAD matches or exceeds leading baselines in efficiency, scalability, and Pareto front coverage.

Citations

• Preference-Guided Diffusion for Multi-Objective Offline Optimization
• Multi-objective methods in Federated Learning: A survey and taxonomy
• Improving Pareto Set Learning for Expensive Multi-objective Optimization via Stein Variational Hypernetworks
• ParetoFlow: Guided Flows in Multi-Objective Optimization
• Multi-objective Deep Learning: Taxonomy and Survey of the State of the Art
• LibMOON: A Gradient-based MultiObjective OptimizatioN Library in PyTorch
• Offline Multi-Objective Optimization
• Expensive Multi-Objective Bayesian Optimization Based on Diffusion Models
• Diffusion Model for Data-Driven Black-Box Optimization
• Smooth Tchebycheff Scalarization for Multi-Objective Optimization
• PMGDA: A Preference-based Multiple Gradient Descent Algorithm
• qPOTS: Efficient batch multiobjective Bayesian optimization via Pareto optimal Thompson sampling
• Importance-aware Co-teaching for Offline Model-based Optimization
• Parallel-mentoring for Offline Model-based Optimization
• Diffusion Models for Black-Box Optimization
• Scalable Diffusion Models with Transformers
• Data-Driven Offline Decision-Making via Invariant Representation Learning
• Pareto Set Learning for Expensive Multi-Objective Optimization
• Classifier-Free Diffusion Guidance
• Diffsound: Discrete Diffusion Model for Text-to-sound Generation
• Equivariant Diffusion for Molecule Generation in 3D
• RoMA: Robust Model Adaptation for Offline Model-based Optimization
• Multi-Objective Bayesian Optimization over High-Dimensional Search Spaces
• Conservative Objective Models for Effective Offline Model-Based Optimization
• Diffusion Models Beat GANs on Image Synthesis
• Improved Denoising Diffusion Probabilistic Models
• Multi-Objective Learning to Predict Pareto Fronts Using Hypervolume Maximization
• DiffWave: A Versatile Diffusion Model for Audio Synthesis
• Denoising Diffusion Probabilistic Models
• Pymoo: Multi-Objective Optimization in Python
• Gradient Surgery for Multi-Task Learning
• Multi-Task Learning as Multi-Objective Optimization
• A Flexible Framework for Multi-Objective Bayesian Optimization using\n Random Scalarizations
• A Comparison of Three Methods for Selecting Values of Input Variables in the Analysis of Output From a Computer Code
• Diffusion-BBO: Diffusion-Based Inverse Modeling for Online Black-Box Optimization

Related

• Fewer Denoising Steps or Cheaper Per-Step Inference: Towards Compute-Optimal Diffusion Model Deployment
• Adaptive Non-uniform Timestep Sampling for Accelerating Diffusion Model Training
• On the Asymptotic Mean Square Error Optimality of Diffusion Models
• You Only Look One Step: Accelerating Backpropagation in Diffusion Sampling with Gradient Shortcuts
• Optimizing DDPM Sampling with Shortcut Fine-Tuning