Minimum-Excess-Work Guidance

2025/05/19 by Christopher Kolloff, Tobias Höppe, Kolloff, Christopher +11 · 1 voice · 1 citation
Computer Science · Mathematics · #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning (stat.ML) #cs.LG #stat.ML

paper · pdf · doi:10.48550/arxiv.2505.13375

arxiv published 2025/05/19 · arxiv updated 2025/05/23

Abstract

We propose a regularization framework inspired by thermodynamic work for guiding pre-trained probability flow generative models (e.g., continuous normalizing flows or diffusion models) by minimizing excess work, a concept rooted in statistical mechanics and with strong conceptual connections to optimal transport. Our approach enables efficient guidance in sparse-data regimes common to scientific applications, where only limited target samples or partial density constraints are available. We introduce two strategies: Path Guidance for sampling rare transition states by concentrating probability mass on user-defined subsets, and Observable Guidance for aligning generated distributions with experimental observables while preserving entropy. We demonstrate the framework's versatility on a coarse-grained protein model, guiding it to sample transition configurations between folded/unfolded states and correct systematic biases using experimental data. The method bridges thermodynamic principles with modern generative architectures, offering a principled, efficient, and physics-inspired alternative to standard fine-tuning in data-scarce domains. Empirical results highlight improved sample efficiency and bias reduction, underscoring its applicability to molecular simulations and beyond.

Citations

• Transferable Generative Models Bridge Femtosecond to Nanosecond Time-Step Molecular Dynamics
• Scalable Equilibrium Sampling with Sequential Boltzmann Generators
• Scalable emulation of protein equilibrium ensembles with generative deep learning
• Thermodynamic Interpolation: A generative approach to molecular thermodynamics and kinetics
• Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Control
• SemlaFlow -- Efficient 3D Molecular Generation with Latent Attention and Equivariant Flow Matching
• D-Flow: Differentiating through Flows for Controlled Generation
• Transition Path Sampling with Boltzmann Generator-based MCMC Moves
• Diffusion Model Alignment Using Direct Preference Optimization
• A Computational Framework for Solving Wasserstein Lagrangian Flows
• SE(3) Equivariant Augmented Coupling Flows
• Equivariant flow matching
• On Kinetic Optimal Probability Paths for Generative Models
• Training Diffusion Models with Reinforcement Learning
• Stochastic Interpolants: A Unifying Framework for Flows and Diffusions
• Aligned Diffusion Schrödinger Bridges
• Universal Guidance for Diffusion Models
• Timewarp: Transferable Acceleration of Molecular Dynamics by Learning Time-Coarsened Dynamics
• Improving and generalizing flow-based generative models with minibatch optimal transport
• Two for One: Diffusion Models and Force Fields for Coarse-Grained Molecular Dynamics
• Action Matching: Learning Stochastic Dynamics from Samples
• Flow Matching for Generative Modeling
• The Vendi Score: A Diversity Evaluation Metric for Machine Learning
• Diffusion Posterior Sampling for General Noisy Inverse Problems
• Rectified Flow: A Marginal Preserving Approach to Optimal Transport
• Classifier-Free Diffusion Guidance
• Torsional Diffusion for Molecular Conformer Generation
• Machine Learning in Molecular Dynamics Simulations of Biomolecular Systems
• The Schrödinger Bridge between Gaussian Measures has a Closed Form
• Denoising Diffusion Restoration Models
• GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models
• Score-Based Generative Modeling through Stochastic Differential Equations
• Denoising Diffusion Probabilistic Models
• Equivariant Flows: Exact Likelihood Generative Learning for Symmetric\n Densities
• TrajectoryNet: A Dynamic Optimal Transport Network for Modeling Cellular\n Dynamics
• Normalizing Flows for Probabilistic Modeling and Inference
• Neural Ordinary Differential Equations
• Attention Is All You Need
• Metainference: A Bayesian Inference Method for Heterogeneous Systems
• Markov models of molecular kinetics: Generation and validation
• Towards a Theory of Transition Paths
• Mapping Long-Range Interactions in α-Synuclein using Spin-Label NMR and Ensemble Molecular Dynamics Simulations
• Simultaneous determination of protein structure and dynamics
• Polar factorization and monotone rearrangement of vector‐valued functions
• A computational fluid mechanics solution to the Monge-Kantorovich mass transfer problem

Cited by

• Calibrating Generative Models

Discussions

• TL;DR: MEW guidance offers a principled way to adjust generated distributions to match experimental data and focus sampling on specific subspaces without the overhead of retraining or heavy data deman [bsky, 2 points, 1 comments]

Related

• Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning
• Energy Matching: Unifying Flow Matching and Energy-Based Models for Generative Modeling
• Efficient Regression-Based Training of Normalizing Flows for Boltzmann Generators
• FALCON: Few-step Accurate Likelihoods for Continuous Flows
• Action-Minimization Meets Generative Modeling: Efficient Transition Path Sampling with the Onsager-Machlup Functional