Beomsu Baek and Mingon Kang (both Computer Science) published a paper, "Stochastic LASSO: enhanced high-dimensional LASSO for high-throughput genomic data," in Scientific Reports. Stochastic LASSO, a powerful linear-based feature selection method designed for extremely high-dimensional, low-sample-size (EHDLSS) genomic data.

Key highlights:

• Stochastic LASSO addresses critical limitations of existing bootstrap-based LASSO methods, including multicollinearity within bootstrap samples, missing predictors due to resampling, and instability from random predictor selection.

• Our method consistently outperforms state-of-the-art LASSO variants, including Hi-LASSO, Recursive, Random, Precision, Relaxed, and Adaptive LASSOs, in multiple experimental settings.