Analysis variation value momentum algorithm backpropagation method in the recognizing process of temperature pattern in Medan

2nd Nommensen International Conference on Technology and Engineering

Putra, Purwa Hasan^1 ; Zarlis, Muhammad^1

Fakultas Ilmu Komputer Dan Teknologi Informasi, University of Sumatera Utara, Medan, Indonesia^1

关键词: Number of iterations;    Temperature patterns;    Training process;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/420/1/012135/pdf
DOI  :  10.1088/1757-899X/420/1/012135

来源: IOP

【 摘 要 】

In this paper, From the results of the study it can be concluded. A different target error will produce a different number of iterations. The smaller the target error, the greater the number of iterations. The smaller the target error, the better the accuracy value (the greater). The number of iterations in the training process with variations in momentum values is smaller than without variations in momentum values. The highest level of accuracy is achieved at the target error of 0.001 with a momentum value of 0.4 which is 96.62%.