Hubert Drazkowski

hubert(dot)drazkowski(at)di.ku.dk

I am

a first year PhD fellow in Theoretical Machine Learning at University of Copenhagen
advised by Yevgeny Seldin and Sebastian Weichwald
a member of CoCaLa and DeLTa labs
interested in causal, sequential and interactive

news

Jun 03, 2026	Attending the Reinforcement Learning Summer School in Milano RLSS 2026.
Apr 15, 2026	I will present my new working paper Identifying Causal Effects using Proximal Time Series: Correcting for the Past in VARs with Proxies at the EuroCIM 2026 Workshop.
Mar 09, 2026	A paper Cross-Sectional Conditional Independence in Stationary Time Series: Graphical Equivalence and Completeness of Collider Separation is accepted for oral presentation at CLeaR 2026 (notable 8.6% of submissions).
Mar 09, 2026	A paper Beware of Hinges in Proximal Variables Regression: Adjusting for Colliding-Mediators with Nuisance PV is accepted at CLeaR 2026.
Dec 06, 2025	EurIPS — poster on extending proximal variables regression (Causality for Impact workshop)

selected publications

CLeaR
Oral

Cross-Sectional Conditional Independence in Stationary Time Series: Graphical Equivalence and Completeness of Collider Separation

Hubert Drazkowski

In Conference on Causal Learning and Reasoning (CLeaR), 2026

Abs PDF Slides

Proximal variables regression estimates effects in the presence of unobserved confounding by using conditionally separated proxies. We study a structural violation caused by variables that act simultaneously as mediators and colliders between them.
CLeaR

Beware of Hinges in Proximal Variables Regression: Adjusting for Colliding-Mediators with Nuisance PV

Hubert Drazkowski

In Conference on Causal Learning and Reasoning (CLeaR), 2026

Abs PDF

We study cross-sectional conditional independence in time series. We reconcile two graphical separations and prove for them strong completeness within GaussianVAR.