Paper accepted to IncrLearn @ ICDM'20

I am happy to share, that our project with Maurits Kaptein and Aurelie Lemmens has been accepted to the IncrLearn workshop at ICDM’20. Main points of the paper:

  • In this paper we describe an algorithm, that connects batch-mode active learning and individualized treatment effect (ITE) estimation
  • We are utilizing BART’s predictive uncertainty to select informative units.
  • We show by extensive simulations, that B-EMCMITE can outperform other selection strategies in an one-shot setup.

If you have questions about the project, let me know! Paper and related projects will follow soon :)

Update: Presentation is available here

Rock Paper Scissors Vectors by Vecteezy

Zoltán Puha
Zoltán Puha
PhD student in Statistics/Data Science

I am an applied scientist at Zalando, working at the intersection of machine learning and economics. I am currently finishing my PhD dissertation in active learning and causal inference.