arXiv：cs.LG· Tiago da Silva, Amauri H. Souza, Salem Lahlou·· 4 小时前AI 评分33
学习 GFlowNets 的混合模型
Learning a Mixture of GFlowNets
AI 导读
研究者提出描述 GFlowNets 混合模型的通用理论框架，并将其细分为连续索引（CI）与离散索引（DI）两类。CI GFlowNets 可解释为随机特征展开，能提升图结构任务中采样器的表达能力并降低学习不稳定性；DI GFlowNets 则涵盖已有 GFlowNet 训练方法，并支撑新提出的 Stratum-Conditioned（SC）GFlowNets。
正文
Abstract:Learning an ensemble of GFlowNets to sample from a discrete target distribution has become a common approach for achieving better state space exploration and convergence than that of a monolithic sampler. However, these methods often add a substantial runtime overhead to the base model, and their conceptual connection remains elusive. To address this, we first propose a general-purpose theoretical framework for describing a mixture of GFlowNets, which we specialize into continuously (CI) and discretely indexed (DI) collections. On the one hand, we show CI GFlowNets can be interpreted through the lens of a random features expansion, provably boosting the sampler's expressivity in graph-structured tasks and reducing learning instability via spectral shifting. On the other hand, we demonstrate DI GFlowNets encompass prior approaches for GFlowNet training and provide the foundation for the newly proposed Stratum-Conditioned (SC) GFlowNets. This method, which is inspired by the Doob's h-transform of Markov chains, decomposes the state space according to a prescribed modular function and restricts each component to sample from a distinct subset of it. Importantly, SC GFlowNets support centralized and component-wise embarrassingly parallel training, and we show both of them significantly speed up learning convergence and mode coverage without introducing any non-negligible extra computation.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Machine Learning (stat.ML) |
| Cite as: | arXiv:2610.07562 [cs.LG] |
| (or arXiv:2610.07562v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.07562 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Amauri Souza [view email]
[v1]
Tue, 6 Oct 2026 00:49:26 UTC (6,836 KB)
来源：arXiv：cs.LG · arxiv.org