INGI Seminar

Joint optimization of predictive performance and selection stability

by Victor Hamer, INGI PhD student

Feature selection aims at finding the most informative input variables for a prediction task such as classification or regression. Current selection methods, especially applied to high dimensional data, tend to suffer from instability since marginal modifications in the data may result in largely distinct selected feature sets. Such instability strongly limits a sound interpretation of the selected variables by domain experts.

In this talk, I address this issue by optimizing jointly the predictive accuracy and selection stability and by deriving Pareto-optimal trajectories in a bi- objective framework. I will further show that current stability measures are insufficient to make the aforementioned optimization problem sound. More robust alternatives are discussed.

Victor Hamer is a PhD student at the INGI department at the UCLouvain, Belgium. His current work focuses on the realization of a tradeoff between the classical predictive performance of feature selection methods and their selection stability.