Stochastic vs. BFGS Training in Neural Discrimination of RF-Modulation

Abstract

Neuromorphic classification of RF-Modulation type is an on-going topic in SIGINT applications. Neural network training approaches are varied, each being suited to a certain application. For exemplification I show the results for BFGS (Broyden-Fletcher-Goldfarb-Shanno) optimization in discriminating AM vs FM modulation and of stochastic optimization for the challenging case of AM-LSB vs. AM-USB (upper / lower sideband) discrimination. Although slower than BFGS, the stochastic training of a neural network avoids better local minima, obtaining a stable neurocore.