Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

Causal3DIdent dataset

The Causal3DIdent dataset is a synthetic visual benchmark designed for causal representation learning and disentanglement tasks in machine learning. It consists of three-dimensional rendered images featuring various object classes situated in a controlled environment, where ground-truth generative factors—such as object position, orientation, color, lighting, and background characteristics—are governed by an explicit causal graph. By providing high-dimensional image observations alongside known structural relationships and interventions among underlying latent variables, the dataset serves as an evaluation platform to test whether models can reliably discover latent causal mechanisms, disentangle generative factors, and generalize under distributional shifts.

1 item

Weakly supervised causal representation learning

Weakly supervised causal representation learning

Johann Brehmer, Pim de Haan, Phillip Lippe, Taco S. Cohen

OrganizationsQualcommQUVA LabUniversity of Amsterdam

Why you should read this

Proves that high-level causal variables and mechanisms can be identified from pixel-level data paired across unknown interventions, and introduces implicit latent causal models to learn these structures without optimizing discrete graphs.

Learning high-level causal representations together with a causal model from unstructured low-level data such as pixels is impossible from observational data alone. We prove under mild assumptions that this representation is however identifiable in a weakly supervised setting. This involves a dataset with paired samples before and after random, unknown interventions, but no further labels. We then introduce implicit latent causal models, variational autoencoders that represent causal variables and causal structure without having to optimize an explicit discrete graph structure. On simple image data, including a novel dataset of simulated robotic manipulation, we demonstrate that such models can reliably identify the causal structure and disentangle causal variables.

Added

2026-09-30