Performance analysis of various training algorithms of deep learning based controller

Abstract Advances in artificial neural networks (ANN), specifically deep learning (DL), have widened the application domain of process control. DL algorithms and models become quite common these days. The training algorithm is most important part an ANN that affects performance controller. Training optimize weights biases according to input-output patterns. In this paper, different was evaluated, analysed, compared a feed-forward backpropagation architecture. were simulated on MATLAB R2021b with license number 1075356. data generated using two benchmark problems control system. performance, gradient, error, validation testing regression obtained analysed. shows Levenberg-Marquardt (LM) produced best value 2.669*10 −14 at 2000 epochs, while ‘traingd’ ‘traingdm’ did not improve beyond their initial values. LM tends produce better results than other algorithms. These indicate suits types problems. also suggest choice can significantly impact network.