Impact of optimizers functions on detection of Melanoma using transfer learning architectures

• DOI: 10.1007/s11042-024-19561-6
• Publication Date: 2024-06-12T03:36:30Z
• Authors: Serhat Kılıçarslan; Hatice Aktas Aydın; Kemal Adem; Esra Kavalcı Yılmaz
• Keywords: Transfer of learning; Contextual image classification

Abstract Early diagnosis-treatment of melanoma is very important because its dangerous nature and rapid spread. When diagnosed correctly early, the recovery rate patients increases significantly. Physical methods are not sufficient for diagnosis classification. The aim this study to use a hybrid method that combines different deep learning in classification investigate effect optimizer used on performance. In study, Melanoma detection was carried out from skin lesions image through simulation created with architectures DenseNet, InceptionV3, ResNet50, InceptionResNetV2 MobileNet seven optimizers: SGD, Adam, RmsProp, AdaDelta, AdaGrad, Adamax Nadam. results show SGD has better more stable performance terms convergence rate, training speed than other optimizers. addition, momentum parameter added structure reduces oscillation time compared functions. It observed best among combined achieved using DenseNet model test accuracy 0.949, sensitivity 0.9403, F score 0.9492.