BSTabDiff: 用于高维表格数据生成的块-子单元扩散先验
BSTabDiff: Block-Subunit Diffusion Priors for High-Dimensional Tabular Data Generation
AI总结：
针对高维低样本量表格数据，提出BSTabDiff框架，通过将特征划分为潜在块并使用共享低维子单元变量生成每个块，结合扩散先验和copula依赖，实现稳定合成与可控基准生成。
摘要：
高维低样本量（HDLSS）表格领域（例如组学）的特点是 $n \ll m$，其中 $n$ = 样本数，$m$ = 特征数。此类领域通常表现出强局部相关组、稀疏跨组依赖、重尾非高斯边缘分布、异方差噪声和结构化缺失，使得在 $\mathbb{R}^m$ 中直接进行密度学习因 $n \ll m$ 而病态。我们提出 BSTabDiff，一种块-子单元生成框架，将 $m$ 个观测特征划分为 $M$ 个潜在块（$M \ll m$），并通过共享的低维子单元变量生成每个块，将全局依赖学习集中在紧凑的块潜在空间 $\mathbb{R}^M$ 中，同时通过 copula 驱动的依赖、灵活的逐特征边缘分布和显式缺失机制解码到完整特征空间。BSTabDiff 支持块潜在上的现代深度先验，包括扩散和归一化流，从而在 HDLSS 场景中实现稳定合成和可控基准生成。实验表明，与 HDLSS 数据上的非结构化表格生成器相比，BSTabDiff 能产生更真实和稳定的高维合成数据。
英文摘要：
High-Dimensional Low-Sample Size (HDLSS) tabular domains (e.g., omics) are characterized by $n \ll m$, where $n$ = number of samples, and $m$ = number of features. Such domains often exhibit strong local correlation groups, sparse cross-group dependencies, heavy-tailed non-Gaussian marginals, heteroscedastic noise, and structured missingness, making direct density learning in $\mathbb{R}^m$ ill-conditioned since $n \ll m$. We propose BSTabDiff, a block-subunit generative framework that partitions the $m$ observed features into $M$ latent blocks ($M \ll m$) and generates each block via a shared low-dimensional subunit variable, concentrating global dependence learning in the compact block-latent space $\mathbb{R}^M$ while decoding to the full feature space with copula-driven dependence, flexible per-feature marginals, and explicit missingness mechanisms. BSTabDiff supports modern deep priors on block latents, including diffusion and normalizing flows, enabling stable synthesis and controllable benchmark generation in the HDLSS regime. Empirically, BSTabDiff produces more realistic and stable high-dimensional synthetic data when compared with unstructured tabular generators on HDLSS data.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
