Abstract:

Quadratic functions are fundamental to understand optimization landscapes in machine learning. In this study, we focus on the optimization of a class of quadratic loss functions, examining their behavior through the construction of Hessian matrices and the analysis of their eigenvalues.