mlsanityData to 7 Oct 2026

Exploring Non-Convex Discrete Energy Landscapes: An Efficient Langevin-Like Sampler with Replica Exchange

arXiv 2501.17323 ↗First posted 28 Jan 2025cs.LG stat.MLcited by 0

Not on the map. Why?

Abstract (arXiv)

Gradient-based Discrete Samplers (GDSs) are effective for sampling discrete energy landscapes. However, they often stagnate in complex, non-convex settings. To improve exploration, we introduce the Discrete Replica EXchangE Langevin (DREXEL) sampler and its variant with Adjusted Metropolis (DREAM). These samplers use two GDSs at different temperatures and step sizes: one focuses on local exploitation, while the other explores broader energy landscapes. When energy differences are significant, sample swaps occur, which are determined by a mechanism tailored for discrete sampling to ensure detailed balance. Theoretically, we prove that the proposed samplers satisfy detailed balance and converge to the target distribution under mild conditions. Experiments across 2d synthetic simulations, sampling from Ising models and restricted Boltzmann machines, and training deep energy-based models further confirm their efficiency in exploring non-convex discrete energy landscapes.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 25 of its 59 references that are in the map. First posted on arXiv 28 Jan 2025.

Builds on

25 of 59 references.

References · 25

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  3. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  4. 2013Efficient Estimation of Word Representations in Vector Spacecited by 4,346
  5. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 712
  6. 2023Hard Prompts Made Easy: Gradient-Based Discrete Optimization for Prompt Tuning and Discoverycited by 329
  7. 2021Deep Generative Modelling: A Comparative Review of VAEs, GANs, Normalizing Flows, Energy-Based and Autoregressive Modelscited by 251
  8. 2019Cyclical Stochastic Gradient MCMC for Bayesian Deep Learningcited by 203
  9. 2015Non-asymptotic convergence analysis for the Unadjusted Langevin Algorithmcited by 195
  10. 2022Score-based Continuous-time Discrete Diffusion Modelscited by 167
  11. 2018Log-concave sampling: Metropolis-Hastings algorithms are fastcited by 117
  12. 2022Generative Flow Networks for Discrete Probabilistic Modelingcited by 113
13 more
  1. 2021Oops I Took A Gradient: Scalable Sampling for Discrete Distributionscited by 92
  2. 2014Consistency and fluctuations for stochastic gradient Langevin dynamicscited by 65
  3. 2020Optimal dimension dependence of the Metropolis-Adjusted Langevin Algorithmcited by 57
  4. 2024Beyond ELBOs: A Large-Scale Evaluation of Variational Methods for Samplingcited by 55
  5. 2016On the Convergence of Stochastic Gradient MCMC Algorithms with High-Order Integratorscited by 35
  6. 2017Beyond Log-concavity: Provable Guarantees for Sampling Multi-modal Distributions using Simulated Tempering Langevin Monte Carlocited by 33
  7. 2019Stochastic Runge-Kutta Accelerates Langevin Monte Carlo and Beyondcited by 33
  8. 2020Non-convex Learning via Replica Exchange Stochastic Gradient MCMCcited by 33
  9. 2020Learning Discrete Energy-based Models via Auxiliary-variable Local Explorationcited by 29
  10. 2020Accelerating Nonconvex Learning via Replica Exchange Langevin Diffusioncited by 15
  11. 2016Stochastic Quasi-Newton Langevin Monte Carlocited by 14
  12. 2020Accelerating Convergence of Replica Exchange Stochastic Gradient MCMC via Variance Reductioncited by 10
  13. 2024Gradient-based Discrete Sampling with Automatic Cyclical Schedulingcited by 9

Built on by

Not cited yet.