Analysis of Customized Optimizers of Convolutional Neural Networks for Lung Cancer Detection

Abstract

Convolutional Neural Network (CNN) is a powerful tool used for classifying medical images. Based on extracted features from CT scan Image CNN classify it as malicious or non-malicious. Optimizers are strategies or methodologies which make a change in the weights of parameters in several iterations and try to minimize losses. Tuning hyperparameters of networks is time consuming and cumbersome task. For training a dataset many customized optimizers and metaheuristic algorithms are available. In this research study, the implementation and analysis of various customized optimizers are done on IQ-OTH/NCCD dataset. Out of six optimizers, Adam reaches 99.84% whereas RmsProp, Nadam and Admax occupied 1.