Date:
12.06.2026, 13:00-13:45
Location:
Zoom (Meeting ID: 675 8938 0001)
Language:
English
Organized by:
IC D2MCM
Register here:
Not required.
Contact:
iz-d2mcm.contact@hu-berlin.de
Correcting inherent bias in iterative imputation of missing data
Joint Seminar Series on Digitality and Digital Methods, 12.06.2026
Speaker: Katarzyna Reluga (WiWiF)
Zoom: Meeting ID: 675 8938 0001
Time: 1-1:45 pm, Fridays
Host: Management team of IZ D2MCM
Contact: iz-d2mcm.contact@hu-berlin.de
Full Programm: Link to Introduction and Programm
Missing data imputation, where a model is trained on observed data to estimate unobserved values, is a central task in machine learning. We show that when the probability of missingness depends on the data, many state-of-the-art methods fail to account for the resulting distribution shift between the observed data used for training and the full data distribution used for evaluation. To address this, we propose a novel imputation algorithm designed to learn an imputation model from the observed data while explicitly accounting for the distribution shift between observed and full data distributions. Simulation studies show consistent improvements over otherwise identical uncorrected baselines, with average reductions of 3\% in RMSE and 7\% in Wasserstein distance.
The format consists of a 30-minute presentation followed by a 15-minute discussion. The presentations are organised via Zoom to enable broad participation. All lectures are also listed in the Event-Section of the website.