arXiv cs.AI· Zheyu Zhang (University of Michigan), Jiayuan Dong (University of Michigan), Jie Liu (University of Michigan), Xun Huan (University of Michigan)·· 4 小时前AI 评分39
GO-CBED：面向因果学习的目标导向序贯贝叶斯实验设计框架
Goal-Oriented Sequential Bayesian Experimental Design for Causal Learning
AI 导读
研究者提出 GO-CBED，一个目标导向的贝叶斯序贯因果实验设计框架，直接最大化用户指定因果量的期望信息增益（EIG），而非推断完整因果模型。该框架通过基于 transformer 的策略网络与 normalizing flow 变分后验联合优化变分下界估计，并借助摊销网络实现实时决策。在合成结构因果模型和半合成基因调控网络等任务中，GO-CBED 在实验预算受限场景下持续优于现有基线。
正文 · 原文
Abstract:We present GO-CBED, a goal-oriented Bayesian framework for sequential causal experimental design. Unlike conventional approaches that select interventions aimed at inferring the full causal model, GO-CBED directly maximizes the expected information gain (EIG) on user-specified causal quantities of interest, enabling more targeted and efficient experimentation. The framework is both non-myopic, optimizing over entire intervention sequences, and goal-oriented, targeting only model aspects relevant to the causal query. To address the intractability of exact EIG computation, we introduce a variational lower bound estimator, optimized jointly through a transformer-based policy network and normalizing flow-based variational posteriors. The resulting policy enables real-time decision-making via an amortized network. We demonstrate that GO-CBED consistently outperforms existing baselines across various causal reasoning and discovery tasks-including synthetic structural causal models and semi-synthetic gene regulatory networks-particularly in settings with limited experimental budgets and complex causal mechanisms. Our results highlight the benefits of aligning experimental design objectives with specific research goals and of forward-looking sequential planning.
| Comments: | Published in Transactions on Machine Learning Research (TMLR), 09/2026. 38 pages, 26 figures, 5 tables |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Methodology (stat.ME); Machine Learning (stat.ML) |
| Cite as: | arXiv:2507.07359 [cs.LG] |
| (or arXiv:2507.07359v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2507.07359 arXiv-issued DOI via DataCite |
|
| Journal reference: | Transactions on Machine Learning Research, 2026 |
Submission history
From: Zheyu Zhang [view email]
[v1]
Thu, 10 Jul 2025 00:53:57 UTC (8,076 KB)
[v2]
Wed, 4 Nov 2026 03:16:48 UTC (8,461 KB)
来源：arXiv cs.AI · arxiv.org