3 papers
cs.LG2025
Making Robust Generalizers Less Rigid with Loss Concentration
Matthew J. Holland, Toma Hamada
While the traditional formulation of machine learning tasks is in terms of performance on average, in practice we are often interested in how well a trained model performs on rare…
stat.ML2024
Soft ascent-descent as a stable and flexible alternative to flooding
Matthew J. Holland, Kosuke Nakatani
As a heuristic for improving test accuracy in classification, the "flooding" method proposed by Ishida et al. (2020) sets a threshold for the average surrogate loss at training tim…
stat.ML2024
Criterion Collapse and Loss Distribution Control
Matthew J. Holland
In this work, we consider the notion of "criterion collapse," in which optimization of one metric implies optimality in another, with a particular focus on conditions for collapse…