Second-OrderOptimalityinNon-Convex DecentralizedOptimizationviaPerturbedGradient Tracking

Simultaneously, convex formulations for training machine learning tasks have been replaced by nonconvex representations such as neural networks.