global minima
Global minima refers to the lowest point in the loss landscape of a function across the entire parameter space. Finding the global minimum is a key objective in optimization tasks in AI, as it corresponds to the best performance of a model.
- A Tale of Two Symmetries: Exploring the Loss Landscape of Equivariant Models
- Global Minimizers of $\ell^p$-Regularized Objectives Yield the Sparsest ReLU Neural Networks
- Neural Collapse under Gradient Flow on Shallow ReLU Networks for Orthogonally Separable Data
- Non-Singularity of the Gradient Descent Map for Neural Networks with Piecewise Analytic Activations
- The Nuclear Route: Sharp Asymptotics of ERM in Overparameterized Quadratic Networks