PulseAugur
EN
LIVE 05:43:19

New method uses Wasserstein distance for ICA and causal inference

By PulseAugur Editorial · [2 sources] ·

Researchers have developed a novel approach to Independent Component Analysis (ICA) and causal inference using the squared 2-Wasserstein distance to the standard Gaussian distribution. This method effectively identifies the unmixing matrix in ICA and characterizes causal orders in Linear Non-Gaussian Acyclic Models (LiNGAM). The study introduces empirical estimators with uniform convergence bounds and presents three practical solvers for ICA, causal-order search, and a greedy variant, demonstrating competitive performance in empirical evaluations. AI

IMPACT Introduces a novel statistical technique that could enhance machine learning capabilities in signal separation and causal discovery.

RANK_REASON The cluster contains a research paper detailing a new statistical method for ICA and causal inference. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New method uses Wasserstein distance for ICA and causal inference

COVERAGE [2]

  1. arXiv stat.ML TIER_1 English(EN) · F\'elix Laplante, Christophe Ambroise, Pierre Humbert ·

    Contrast-Free ICA and Causal Inference via Wasserstein Distances to the Gaussian

    arXiv:2607.12832v1 Announce Type: new Abstract: We study the squared $2$-Wasserstein distance to the standard Gaussian as a non-Gaussianity criterion and use it for linear Independent Component Analysis (ICA) and causal inference in Linear Non-Gaussian Acyclic Models (LiNGAM). Th…

  2. arXiv stat.ML TIER_1 English(EN) · Pierre Humbert ·

    Contrast-Free ICA and Causal Inference via Wasserstein Distances to the Gaussian

    We study the squared $2$-Wasserstein distance to the standard Gaussian as a non-Gaussianity criterion and use it for linear Independent Component Analysis (ICA) and causal inference in Linear Non-Gaussian Acyclic Models (LiNGAM). The analysis relies on a strict inequality between…

RELATED ENTITIES

RELATED TOPICS