Charles Hermite

New theory explains incremental learning in shallow neural networks

Researchers have developed a new theoretical framework for understanding incremental learning in shallow neural networks. This work focuses on polynomial-width two-layer networks trained on orthogonal multi-index target…