説明
Causal Inference Team (https://www.riken.jp/en/research/labs/aip/generic_tech/cause_infer/) at RIKEN AIP
Speaker 1: (10 min) Shohei SHIMIZU:
Title: Overview of the Causal Inference Team
Abstract:
The causal inference team consists of methodologists and scientists and aims to develop data-driven statistical methods for learning causal structures from theoretical and philosophical perspectives. In this talk, I briefly give an overview of the causal inference team.
Speaker 2: (30 min) Takashi Nicholas MAEDA:
Title: Causal discovery in the presence of unobserved variables
Abstract: Causal discovery methods are aimed at inferring causal relations between observed variables. Most of the existing methods assume the absence of unobserved variables. However, in most cases, these assumptions are rarely met. In this talk, I will introduce recent studies on causal discovery in the presence of unobserved variables.
Speaker 3: (30 min) Yan ZENG:
Title: Causal discovery with multi-domain LiNGAM for latent factors
Abstract:
Discovering causal structures among latent factors from observed data is a particularly challenging problem, in which many empirical researchers are interested. Despite its success in certain degrees, existing methods focus on the single-domain observed data only, while in many scenarios data may be originated from distinct domains, e.g. in neuroinformatics. In this talk, we propose Multi-Domain Linear Non-Gaussian Acyclic Models for LAtent Factors (abbreviated as MD-LiNA model) to identify the underlying causal structure between latent factors (of interest), tackling not only single-domain observed data but multiple-domain ones, and provide its identification results. In particular, we first locate the latent factors and estimate the factor loadings matrix for each domain separately. Then to estimate the structure among latent factors (of interest), we derive a score function based on the characterization of independence relations between external influences and the dependence relations between multiple-domain latent factors and latent factors of interest, enforcing acyclicity, sparsity, and elastic net constraints. The resulting optimization thus produces asymptotically correct results. It also exhibits satisfactory capability in regimes of small sample sizes or highly-correlated variables and simultaneously estimates the causal directions and effects between latent factors. Experimental results on both synthetic and real-world data demonstrate the efficacy of our approach. This work will appear in IJCAI2021.
Speaker 4: (30 min) Jun OTSUKA:
Title: Causal modeling from a philosophical perspective
Abstract:
Causal modeling is distinguished from other probabilistic/machine learning methods in that it aims to model change in probability states induced by hypothetical interventions, rather than states themselves. Modeling such counterfactual transitions calls for stronger assumptions than those of conventional statistical models. In this talk I characterize this difference as one concerning underlying ontology: that is, statisticians and causal modelers see the same target phenomenon as different kinds of entities. In particular, causal modeling adopts a stronger ontological assumption (that is, assumes more “things”) than that of statistical modeling, which allows for stronger (counterfactual) reasoning but also presents a harder epistemological challenge in estimating the causal structure. After the philosophical discussion I will introduce, if time permitted, our recent work on the nature of such causal kinds, in particular the formal identity criteria for determining when two causal models represent the same “thing.”
