September 2, 2021 11:00


Title: Post-selection Inference with HSIC-Lasso (ICML 2021)

Detecting influential features in non-linear and/or high-dimensional data is an increasingly important task in machine learning. However, inference on the chosen features can be significantly flawed when the selection procedure is not accounted for. We propose a post-selection inference procedure using the so-called model-free “HSIC-Lasso” based on the framework of truncated Gaussians combined with the polyhedral lemma. We then develop an algorithm, which allows for low computational costs and provides a selection of the regularisation parameter. The performance of the proposed method is illustrated by both artificial and real-world data based experiments, which exhibit a tight control of the type-I error, even for small sample sizes.

Tobias Freidling obtained a Bachelor and Master degree in Mathematics from the Ludwig-Maximilian University and Technical University of Munich, respectively. From March to August 2020 he was a research student at Prof. Yamada’s group at Kyoto University and worked on the project presented in this talk. Since October 2020, Tobias is a PhD student at the University of Cambridge, supervised by Qingyuan Zhao. His research focuses on causal inference, sensitivity analysis and selective inference.

Related Laboratories

last updated on September 10, 2021 14:02Laboratory