Extension of inertial Levenberg–Marquardt convergence results to generalized derivative operators
Let σ:R→R\sigma:\mathbb{R}\to\mathbb{R}σ:R→R be the activation function in the neural-network inverse problem, and let s(t)=∂+σ(t)s(t)=\partial_+\sigma(t)s(t)=∂+σ(t) be its right derivative. Define…