Characterizing Implicit Bias in Terms of Optimization Geometry
[edit]
Proceedings of the 35th International Conference on Machine Learning, PMLR 80:18321841, 2018.
Abstract
We study the bias of generic optimization methods, including Mirror Descent, Natural Gradient Descent and Steepest Descent with respect to different potentials and norms, when optimizing underdetermined linear models or separable linear classification problems. We ask the question of whether the global minimum (among the many possible global minima) reached by optimization can be characterized in terms of the potential or norm, and indecently of hyperparameter choices, such as stepsize and momentum.
Related Material


