Paper ID: 2202.11141
Nonconvex Extension of Generalized Huber Loss for Robust Learning and Pseudo-Mode Statistics
Kaan Gokcesu, Hakan Gokcesu
We propose an extended generalization of the pseudo Huber loss formulation. We show that using the log-exp transform together with the logistic function, we can create a loss which combines the desirable properties of the strictly convex losses with robust loss functions. With this formulation, we show that a linear convergence algorithm can be utilized to find a minimizer. We further discuss the creation of a quasi-convex composite loss and provide a derivative-free exponential convergence rate algorithm.
Submitted: Feb 22, 2022