---
格式版本: 2
标题: "Hamiltonian dynamics for sampling on discrete spaces"
原文链接: "https://arxiv.org/abs/2608.17961"
发布日期: "2026-08-18"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Tue, 18 Aug 2026 16:15:23 UTC (1,083 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-19T21:16:51+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-19T21:16:38+08:00"
入库时间: "2026-08-19T13:16:51.578Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=Scale-up&searchtype=all"
匹配关键词:
  - "Scale-up"
相关厂家:
  []
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 5
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "该页面为arXiv搜索入口页，实际论文是关于离散空间哈密顿采样的统计理论，与超节点、AI Rack、机柜级AI基础设施等主题完全无关，属于搜索词误命中。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-19T21:16:56+08:00"
AI主题相关性: 0
AI来源权威性: 5
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
AI摘要: "Raphaël Barboni等人在离散状态空间上提出一类非可逆Hamiltonian Monte Carlo动力学，通过引入连续动量变量且无需连续嵌入，保证目标分布不变性，并分析遍历性、渐近方差等性质。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:21:15.487Z"
采集批次: "2026年8月19日19点33分50秒"
采集批次ID: "20260819-193350-682"
去重键: "https://arxiv.org/abs/2608.17961"
---

## Mathematics > Statistics Theory

## Title:Hamiltonian dynamics for sampling on discrete spaces

Authors:[Raphaël Barboni](https://arxiv.org/search/math?searchtype=author&query=Barboni,+R) (1), [Sebastiano Grazzi](https://arxiv.org/search/math?searchtype=author&query=Grazzi,+S) (1), [Giacomo Zanella](https://arxiv.org/search/math?searchtype=author&query=Zanella,+G) (1) ((1) Bocconi University)

[View PDF](https://arxiv.org/pdf/2608.17961) [HTML (experimental)](https://arxiv.org/html/2608.17961v1)

> Abstract:We develop a general class of non-reversible Hamiltonian Monte Carlo dynamics on discrete state spaces. The method augments the discrete state with a continuous momentum variable and does not require a continuous embedding of the discrete state space. We establish conditions for invariance of the target distribution and study the resulting processes in terms of ergodicity, exponential contractivity, asymptotic variance and relaxation time. We then derive scaling limits on increasingly fine lattices and high-dimensional hypercubes. In both settings, suitable rescalings converge to Hamiltonian dynamics in continuous space, revealing a diffusive-to-ballistic speed-up over reversible samplers, even for heterogeneous target distributions where standard non-reversible methods become diffusive. The proposed dynamics can be simulated exactly in continuous time, given access to the target distribution at all neighbours of the current state. To reduce computational cost, we provide approximation schemes based on splitting, $\tau$ -leaping and gradient approximations. We illustrate the proposed framework numerically on examples involving high-dimensional hypercubes, mixture models and permutations.

| Subjects: | Statistics Theory (math.ST) |
| --- | --- |
| MSC classes: | 60J27, 60J28, 65C05 (Primary) 60H10, 60H30 (Secondary) |
| Cite as: | [arXiv:2608.17961](https://arxiv.org/abs/2608.17961) \[math.ST\] |
|  | (or [arXiv:2608.17961v1](https://arxiv.org/abs/2608.17961v1) \[math.ST\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.17961](https://doi.org/10.48550/arXiv.2608.17961) |

## Submission history

From: Raphaël Barboni \[[view email](https://arxiv.org/show-email/d6c5092f/2608.17961)\]  
**\[v1\]** Tue, 18 Aug 2026 16:15:23 UTC (1,083 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.17961) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))