文章基本信息

标题：Comparative Performance of Several RecentSupervised Learning Algorithms
本地全文：下载
作者：Tony Bazzi ; Rana Ismail ; Mohamed Zohdy 等
期刊名称：International Journal of Computer and Information Technology
印刷版ISSN：2279-0764
出版年度：2018
卷号：7
期号：2
页码：49
出版社：International Journal of Computer and Information Technology
摘要：A wide variety of optimization algorithms havebeen developed, however their performance is still unclearacross optimization landscapes. The manuscript presentedherein discusses methods for modeling and training neuralnetworks on a small dataset. The algorithms includeconventional gradient descent, Levenberg-Marquardt,Momentum, Nesterov Momentum, ADAgrad, andRMSprop learning methodologies. The work aims tocompare the performance, efficiency, and accuracy of thedifferent algorithms utilizing the fertility dataset availablethrough the UC Irvine machine learning repository.
关键词：Neural Networks; Back Propagation; LebenbergMarquardt;Momentum; Nesterov; ADAgrad; RMSprop;Hyperparameters; Newton Method; Supervised Learning; gradient;descent; delta rule; Python; Fertility