Comparison of variable learning rate and Levenberg-Marquardt back-propagation training algorithms for detecting attacks in Intrusion Detection Systems

摘要

This paper investigates the use of variable learning rate back-propagation algorithm and Levenberg-Marquardt back-propagation algorithm in Intrusion detection system for detecting attacks. In the present study, these 2 neural network (NN) algorithms are compared according to their speed, accuracy and, performance using mean squared error (MSE) (Closer the value of MSE to 0, higher will be the performance). Based on the study and test results, the Levenberg-Marquardt algorithm has been found to be faster and having more accuracy and performance than variable learning rate backpropagation algorithm.