ACTIVA: Amortized Causal Effect Estimation via Transformer-based Variational Autoencoder
Not on the map. Why?
Abstract (arXiv)
Predicting post-intervention distributions from observational data is central to many scientific and decision-making problems, but remains challenging due to causal ambiguity, restrictive modeling assumptions, and the lack of amortization across tasks. We introduce ACTIVA, a transformer-based conditional variational autoencoder for amortized estimation of full interventional distributions from observational data and intervention queries. ACTIVA learns a conditional latent prior that supports zero-shot inference by amortizing causal knowledge across diverse training tasks. We provide a consistency result showing that, under idealized conditions, ACTIVA's learning objective targets a mixture over the interventional distributions of causal models that are observationally compatible with the input. Empirically, on synthetic datasets and biologically realistic gene-expression simulations, ACTIVA substantially outperforms a correlational baseline, reduces spurious non-descendant effects, and achieves competitive performance relative to strong amortized baselines. Our results show that ACTIVA is a promising approach for estimating interventional distributions from observational data.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 22 of its 49 references that are in the map. First posted on arXiv 3 Mar 2025.
Builds on
22 of 49 references.
References · 22
- 2017Attention Is All You Needcited by 59,932
- 2017Causal Effect Inference with Deep Latent-Variable Modelscited by 408
- 2018Inference Suboptimality in Variational Autoencoderscited by 131
- 2021Self-Attention Between Datapoints: Going Beyond Individual Input-Output Pairs in Deep Learningcited by 112
- 2021DiBS: Differentiable Bayesian Structure Learningcited by 105
- 2020Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Datacited by 99
- 2022Amortized Inference for Causal Structure Learningcited by 89
- 2022Diffusion Causal Models for Counterfactual Estimationcited by 87
- 2021Relating Graph Neural Networks to Structural Causal Modelscited by 39
- 2020Supervised Whole DAG Causal Discoverycited by 19
- 2023Zero-shot causal learningcited by 17
- 2024Amortized Active Causal Induction with Deep Reinforcement Learningcited by 10
10 more
- 2024Demystifying amortized causal discovery with transformerscited by 10
- 2024Modular Learning of Deep Causal Generative Models for High-dimensional Causal Inferencecited by 9
- 2021β-Intact-VAE: Identifying and Estimating Causal Effects under Limited Overlapcited by 8
- 2024Learning Structural Causal Models through Deep Generative Models: Methods, Guarantees, and Challengescited by 8
- 2020Targeted VAE: Variational and Targeted Learning for Causal Inferencecited by 7
- 2024CORE: Towards Scalable and Efficient Causal Discovery with Reinforcement Learningcited by 6
- 2023Causal Inference for Banking Finance and Insurance A Surveycited by 5
- 2024CausalPlayground: Addressing Data-Generation Requirements in Cutting-Edge Causality Researchcited by 4
- 2023CMVAE: Causal Meta VAE for Unsupervised Meta-Learningcited by 3
- 2024Estimating Joint Interventional Distributions from Marginal Interventional Datacited by 2
Built on by
Citing papers per year: 2025: 1.
1 paper.