Correlation Analysis Between Error Rate of Output and Learning Rate in Backpropagation Network

Mustafidah, Hindayati and Suwarsito, Suwarsito (2018) Correlation Analysis Between Error Rate of Output and Learning Rate in Backpropagation Network. Advanced Science Letters, 24 (12). pp. 9182-9185. ISSN 1936-7317

Text 1. ASL 24(12) Des 2018 Correlation analysis.pdf Download (547kB)

Abstract

Backpropagation network is the most widely used model of neural network. It is based on some information obtained from Levenberg-Marquardt algorithm which is a learning algorithm in backpropagation network that produces the smallest error rate. Calculation error rate is based on one of the network parameters which is the learning rate. In this research, the analysis was performed to measure the correlation between the level of network errors and learning rate. The analysis began with the development of program code to run a learning algorithm that generated error rates which were based on the variation value of 0.01 to 1.0. The results in the form of error rate and learning rate were analyzed using correlation analysis. From the analysis result, it was concluded that no significant correlation found between the level of errors generated by the network and the learning rate at 5% α.

Actions (login required)

View Item