发表机构
University of Cambridge; Karlsruhe Institute of Technology; MIT; Microsoft Research New England（剑桥大学; 卡尔斯鲁厄理工学院; 麻省理工学院; 微软研究院新英格兰分院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出离散吉布斯迭代神经采样器，一种定点神经采样器，解决现有离散神经采样器的模式崩溃、无收敛保证和依赖特定参考过程等问题，实现高效可扩展学习，并扩展到高维系统和合金相图估计。
AI 中文摘要
在没有数据的情况下从离散、未归一化的分布中采样是一个具有挑战性的问题。神经采样器通过直接从密度评估中训练生成模型提供了一种有前景的方法。尽管最近取得了进展，现有的离散神经采样器容易发生模式崩溃，通过定点迭代训练时没有收敛保证，并且通常与特定的参考过程（如掩码或均匀扩散）绑定。在这项工作中，我们引入了离散吉布斯迭代神经采样器，这是一种定点神经采样器，解决了这些局限性，实现了高效、可扩展的学习，在实践中大幅减少了模式崩溃。我们的框架建立在掩码扩散之上，并扩展到分布对之间的传输。我们证明了所提出的方法能有效扩展到高维系统，支持跨不同条件的摊销采样，并能够准确估计合金相图。
英文摘要
Sampling from discrete, unnormalized distributions without access to data is a challenging problem. Neural samplers offer a promising approach by training generative models from density evaluations directly. Despite recent progress, existing discrete neural samplers are prone to mode collapse, come without convergence guarantees when trained via fixed-point iterations, and are often tied to a specific reference process such as masked or uniform diffusion. In this work, we introduce Discrete Gibbs Iterative Neural Sampler, a fixed-point neural sampler that addresses these limitations, enabling efficient, scalable learning, substantially reducing mode collapse in practice. Our framework builds on masked diffusion and also extends to transport between pairs of distributions. We demonstrate that the resulting method scales effectively to high-dimensional systems, supports amortized sampling across different conditions, and enables accurate estimation of alloy phase diagrams.
