Conditional Flow Matching for ML-Based Inverse Design Problems

📄 arXiv: 2609.00863v1 📥 PDF

作者: Juliana Felder, Milad Habibi, Soheyl Massoudi, Mark Fuge

分类: cs.LG

发布日期: 2026-09-01

备注: 13 pages, 2 figures, 6 tables. Accepted for presentation at EngOpt 2026


💡 一句话要点

提出条件流匹配以解决基于机器学习的逆向设计问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture)

关键词: 逆向设计 条件流匹配 深度生成模型 优化算法 机器学习

📋 核心要点

  1. 现有的优化方法在处理受偏微分方程约束的设计问题时,计算成本高且对初始化敏感,限制了其应用。
  2. 本文提出条件流匹配(CFM)作为一种新方法,能够在推理时快速生成候选设计,减少对计算资源的需求。
  3. 实验结果表明,CFM在多个基准任务中实现了显著的性能提升,尤其在样本生成速度和设计质量方面优于现有方法。

📝 摘要(中文)

工程逆向设计常常受到优化问题的迭代求解器高计算成本和对初始化敏感性的限制。深度生成模型能够在推理时无需重新运行模拟器生成候选设计。本文将条件流匹配(CFM)添加到EngiOpt中,并与条件扩散模型和条件生成对抗网络(cGAN)在EngiBench的结构(beams2d)和热(heatconduction2d)基准上进行比较。通过累积最优性间隙(COG)和最终最优性间隙(FOG)评估生成设计的效果。结果显示,CFM在两个任务上均实现了最低的COG、FOG、最大均值差异(MMD)和体积分数偏差,表现优于其他基线。

🔬 方法详解

问题定义:本文旨在解决基于机器学习的逆向设计问题,现有方法在优化过程中面临高计算成本和对初始化的敏感性,限制了其有效性。

核心思路:论文提出条件流匹配(CFM)作为一种新颖的生成模型,能够在推理阶段快速生成设计候选,避免了传统方法的迭代计算。

技术框架:CFM集成于EngiOpt框架中,采用与条件扩散模型和cGAN相同的下游优化协议,主要模块包括设计生成、性能评估和优化反馈。

关键创新:CFM的核心创新在于其高效的设计生成能力,能够在一次前向传递中生成候选设计,与需要迭代反向积分的扩散模型形成鲜明对比。

关键设计:在实验中,CFM的参数设置经过优化,采用了特定的损失函数以最小化COG和FOG,同时设计了适应性强的网络结构以提高生成效率。

🖼️ 关键图片

fig_0
img_1
img_2

📊 实验亮点

实验结果显示,CFM在beams2d和heatconduction2d任务上分别实现了0.4%和1.0%的均值体积分数偏差,相较于扩散模型的3.8%和11.2%有显著提升。此外,CFM在beams2d任务中达到了53.2样本/秒的生成速度,约为扩散基线的66倍,展现了其优越的性能。

🎯 应用场景

该研究的潜在应用领域包括工程设计、材料科学和建筑优化等,能够显著提高设计过程的效率和质量。通过快速生成高质量的设计候选,CFM有助于加速创新和降低开发成本,未来可能在智能制造和自动化设计中发挥重要作用。

📄 摘要(原文)

Engineering inverse design is often limited by the high computational cost of iterative solvers for optimization problems constrained by partial differential equations (PDEs) and by their sensitivity to initialization. Deep generative models can produce candidate designs without rerunning the simulator at inference time. Generative adversarial networks (GANs) sample in one forward pass, whereas diffusion models require iterative reverse-time integration. In this work, we add conditional flow matching (CFM) to EngiOpt and compare it with a conditional diffusion model and a conditional generative adversarial network (cGAN) on structural (beams2d) and thermal (heatconduction2d) benchmarks from EngiBench using the same downstream optimization protocol. We use cumulative optimality gap (COG) and final optimality gap (FOG) as the primary metrics for evaluating the generated designs as warm starts for gradient-based refinement. On the evaluated EngiOpt implementations and two EngiBench tasks, CFM achieves the lowest measured COG, FOG, maximum mean discrepancy (MMD), and volume-fraction deviation on both tasks. CFM has mean volume-fraction deviations of 0.4% and 1.0% on beams2d and heatconduction2d, respectively, compared with 3.8% and 11.2% for diffusion. At Euler s = 16, CFM achieves 53.2 samples/s on beams2d, about 66 times the measured throughput of the evaluated diffusion baseline using 1000 network evaluations under the same timing protocol, with COG 1.182 +/- 3.126, compared with 1.173 +/- 3.100 for Euler s = 32. Across the two tasks, CFM produces warm starts with lower measured COG than both baselines and uses fewer network evaluations than diffusion.