When Tux Meets the Table: Adventures in Tabular Data Learning

Abstract

This talk explores the landscape of tabular data through a unique and challenging case: the vast configuration space of the Linux kernel. Using the TuxKConfig dataset, we examine core concepts of tabular data, comparing modern approaches across data exploration, feature engineering, classical machine learning and deep learning models, evaluation, and benchmarking. The discussion highlights the inherent challenges of high-dimensional tabular spaces and shows how the TuxKConfig case reveals that tabular learning is more than a set of tools—it is a lens for understanding and extracting insights from complex, configurable systems.