为了有效研究复杂因果系统，通过丢弃无关细节同时保留关键特征来构建系统各部分的抽象通常很有用。信息瓶颈（Information Bottleneck，IB）方法是一种被广泛使用的变量抽象构建方法，它通过压缩随机变量，同时保留对目标变量的预测能力。传统方法（如 IB）纯粹是统计性的，忽视了潜在的因果结构，因此不适合因果任务。我们提出因果信息瓶颈（Causal Information Bottleneck，CIB），这是 IB 的一种因果扩展，它压缩一组选定变量，同时维持对目标变量的因果控制。该方法生成的（一组）变量抽象具有因果可解释性，能让我们洞察被抽象变量与目标变量之间的相互作用，并可用于关于干预的推理。我们给出的实验结果表明，所学习到的抽象能够按预期准确捕捉因果关系。
英文摘要：
To effectively study complex causal systems, it is often useful to construct abstractions of parts of the system by discarding irrelevant details while preserving key features. The Information Bottleneck (IB) method is a widely used approach to construct variable abstractions by compressing random variables while retaining predictive power over a target variable. Traditional methods like IB are purely statistical and ignore underlying causal structures, making them ill-suited for causal tasks. We propose the Causal Information Bottleneck (CIB), a causal extension of the IB, which compresses a set of chosen variables while maintaining causal control over a target variable. This method produces abstractions of (sets of) variables which are causally interpretable, give us insight about the interactions between the abstracted variables and the target variable, and can be used when reasoning about interventions. We present experimental results demonstrating that the learned abstractions accurately capture causal relations as intended.
McGill University（麦吉尔大学）；Facebook AI Research（Facebook人工智能研究院）；Carnegie Mellon University（卡内基梅隆大学）；Purdue University（普渡大学）；California Institute of Technology（加州理工学院）；University of Southern California（南加州大学）；HK3 Lab（HK3实验室）
Intelligent agents can cope with sensory-rich environments by learning task-agnostic state abstractions. In this paper, we propose an algorithm to approximate causal states, which are the coarsest partition of the joint history of actions and observations in partially-observable Markov decision processes (POMDP). Our method learns approximate causal state representations from RNNs trained to predict subsequent observations given the history. We demonstrate that these learned state representations are useful for learning policies efficiently in reinforcement learning problems with rich observat…
School of Software, Shandong University（山东大学软件学院）；SDU-NTU Joint Centre for AI Research, Shandong University（山东大学-南洋理工大学人工智能联合研究中心）；Chongqing Uni. of Posts and Telecom.（重庆邮电大学）；MOE Key Laboratory of Big Data Intelligent Computing, Chongqing Uni. of Posts and Telecom.（重庆邮电大学大数据智能计算教育部重点实验室）
Structural causal models (SCMs) are a widespread formalism to deal with causal systems. A recent direction of research has considered the problem of relating formally SCMs at different levels of abstraction, by defining maps between SCMs and imposing a requirement of interventional consistency. This paper offers a review of the solutions proposed so far, focusing on the formal properties of a map between SCMs, and highlighting the different layers (structural, distributional) at which these properties may be enforced. This allows us to distinguish families of abstractions that may or may not b…
Department of Technology, Management and Economics, Danish Technical University（丹麦技术大学技术、管理与经济学系）；Department of Mathematics and Statistics, Binghamton University, State University of New York（纽约州立大学宾汉姆顿分校数学与统计系）；Department of Mathematical Sciences, University of Copenhagen（哥本哈根大学数学科学系）
Structural analysis methods (e.g., probing and feature attribution) are increasingly important tools for neural network analysis. We propose a new structural analysis method grounded in a formal theory of causal abstraction that provides rich characterizations of model-internal representations and their roles in input/output behavior. In this method, neural representations are aligned with variables in interpretable causal models, and then interchange interventions are used to experimentally verify that the neural representations have the causal properties of their aligned variables. We apply …
The University of Texas at Austin（德克萨斯大学奥斯汀分校）；Sony AI（索尼人工智能）
Learning dynamics models accurately is an important goal for Model-Based Reinforcement Learning (MBRL), but most MBRL methods learn a dense dynamics model which is vulnerable to spurious correlations and therefore generalizes poorly to unseen states. In this paper, we introduce Causal Dynamics Learning for Task-Independent State Abstraction (CDL), which first learns a theoretically proved causal dynamics model that removes unnecessary dependencies between state variables and the action, thus generalizing well to unseen states. A state abstraction can then be derived from the learned dynamics, …
McGill University（麦吉尔大学）；Mila（米拉研究所）；Facebook AI Research（脸书人工智能研究院）；University of Oxford（牛津大学）；Deepmind（深度思维）
Generalization across environments is critical to the successful application of reinforcement learning algorithms to real-world challenges. In this paper, we consider the problem of learning abstractions that generalize in block MDPs, families of environments with a shared latent state space and dynamics structure over that latent space, but varying observations. We leverage tools from causal inference to propose a method of invariant prediction to learn model-irrelevance state abstractions (MISA) that generalize to novel observations in the multi-environment setting. We prove that for certain…
University of Strathclyde（思克莱德大学）；University of Copenhagen（哥本哈根大学）
Interventional causal models describe several joint distributions over some variables used to describe a system, one for each intervention setting. They provide a formal recipe for how to move between the different joint distributions and make predictions about the variables upon intervening on the system. Yet, it is difficult to formalise how we may change the underlying variables used to describe the system, say moving from fine-grained to coarse-grained variables. Here, we argue that compositionality is a desideratum for such model transformations and the associated errors: When abstracting…
Università della Svizzera Italiana（瑞士意大利大学）；University of Liechtenstein（利赫泰森大学）；Politecnico di Torino（托尼诺理工学院）；IBM Research（IBM研究院）；Scuola Universitaria Professionale della Svizzera Italiana（瑞士意大利专业大学）
Kyoto University（京都大学）；Nagahama Institute of Bio-Science and Technology（长滨生物科学技术大学）
We develop a category-theoretic criterion for determining the equivalence of causal models having different but homomorphic directed acyclic graphs over discrete variables. Following Jacobs et al. (2019), we define a causal model as a probabilistic interpretation of a causal string diagram, i.e., a functor from the ``syntactic'' category $\textsf{Syn}_G$ of graph $G$ to the category $\textsf{Stoch}$ of finite sets and stochastic matrices. The equivalence of causal models is then defined in terms of a natural transformation or isomorphism between two such functors, which we call a $Φ$-abstracti…
The information bottleneck (IB) method is a feasible defense solution against adversarial attacks in deep learning. However, this method suffers from the spurious correlation, which leads to the limitation of its further improvement of adversarial robustness. In this paper, we incorporate the causal inference into the IB framework to alleviate such a problem. Specifically, we divide the features obtained by the IB method into robust features (content information) and non-robust features (style information) via the instrumental variables to estimate the causal effects. With the utilization of s…
Beijing Institute of Technology（北京理工大学）；University of Aberdeen（阿伯丁大学）
Causal inference is to estimate the causal effect in a causal relationship when intervention is applied. Precisely, in a causal model with binary interventions, i.e., control and treatment, the causal effect is simply the difference between the factual and counterfactual. The difficulty is that the counterfactual may never been obtained which has to be estimated and so the causal effect could only be an estimate. The key challenge for estimating the counterfactual is to identify confounders which effect both outcomes and treatments. A typical approach is to formulate causal inference as a supe…
University of California Los Angeles（加州大学洛杉矶分校）；University College London（伦敦大学学院）；USC Information Sciences Institute（南加州大学信息科学研究所）；Imperial College London（帝国理工学院）；U.S. Army Research Laboratory（美国陆军研究实验室）；U.S. Air Force Research Lab（美国空军研究实验室）
Recent efforts in interpretable deep learning models have shown that concept-based explanation methods achieve competitive accuracy with standard end-to-end models and enable reasoning and intervention about extracted high-level visual concepts from images, e.g., identifying the wing color and beak length for bird-species classification. However, these concept bottleneck models rely on a necessary and sufficient set of predefined concepts-which is intractable for complex tasks such as video classification. For complex tasks, the labels and the relationship between visual elements span many fra…
Learning structured representations of visual scenes is currently a major bottleneck to bridging perception with reasoning. While there has been exciting progress with slot-based models, which learn to segment scenes into sets of objects, learning configurational properties of entire groups of objects is still under-explored. To address this problem, we introduce Constellation, a network that learns relational abstractions of static visual scenes, and generalises these abstractions over sensory particularities, thus offering a potential basis for abstract relational reasoning. We further show …
Istituto Italiano di Tecnologia（意大利理工学院）；Dept. of Neurobiology, Harvard Medical School（哈佛医学院神经生物学系）；Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）；Max Planck ETH Center for Learning Systems（马克斯·普朗克-ETH学习系统中心）；Max Planck Institute for Biological Cybernetics（马克斯·普朗克生物控制论研究所）
The inference of causal relationships using observational data from partially observed multivariate systems with hidden variables is a fundamental question in many scientific domains. Methods extracting causal information from conditional independencies between variables of a system are common tools for this purpose, but are limited in the lack of independencies. To surmount this limitation, we capitalize on the fact that the laws governing the generative mechanisms of a system often result in substructures embodied in the generative functional equation of a variable, which act as sufficient s…
Accuracy and generalization of dynamics models is key to the success of model-based reinforcement learning (MBRL). As the complexity of tasks increases, so does the sample inefficiency of learning accurate dynamics models. However, many complex tasks also exhibit sparsity in the dynamics, i.e., actions have only a local effect on the system dynamics. In this paper, we exploit this property with a causal invariance perspective in the single-task setting, introducing a new type of state abstraction called \textit{model-invariance}. Unlike previous forms of state abstractions, a model-invariance …
Knowledge Abstraction for Knowledge-based Semantic Communication: A Generative Causality Invariant Approach
Department of Intelligent Computing and Data Science, VinUniversity（智能计算与数据科学系，文大学）；School of Computer Science and Statistics, Trinity College Dublin（计算机科学与统计学系，三一学院）；School of Computer Science, The University of Sydney（计算机科学系，悉尼大学）；School of Engineering, The University of Newcastle（工程系，新castle大学）；Department of Information Convergence Engineering, Pusan National University（信息融合工程系，釜山国立大学）
