Improved Run Time for Artificial Neural Networks with Modified Performance of Activation Function in Hidden Layers through Adaptive Learning

  • Mayank Jhalani, Dr. Rakesh Kumar Bhujade

Abstract

Progression in Artificial Neural Network continually plays an essential job in complex example acknowledgment framework. In this article improved execution of the actuation capacities for the hidden layer in fake neural system managed by back propagation calculation with versatile learning highlight has been recorded for design acknowledgment. Multifaceted nature of huge information of example, for example, face acknowledgment, malignancy identification, object acknowledgment, number plate observation and so on is expanding step by step. To determine unpredictability, execution of covered up lair layer is enrolled utilizing Log-sigmoid, ReLU and purelin activation functions individually because of their characteristic properties. A magnificent neural system preparing model of 1350 Alpha-Numeric informational index with 3000 Epoch (cycles) have been prepared in neural network through initiation capacities for number plate acknowledgment. Consequently the exhibition proficiency of hidden layer actuation capacities is recorded for pruning the general back engendering neural system design with improved learning rate alongside better time complexity for design coordinating.

 

Published
2020-05-20
How to Cite
Mayank Jhalani, Dr. Rakesh Kumar Bhujade. (2020). Improved Run Time for Artificial Neural Networks with Modified Performance of Activation Function in Hidden Layers through Adaptive Learning. International Journal of Advanced Science and Technology, 29(06), 6219 - 6231. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/19907