High Impact Factor : 4.396 icon | Submit Manuscript Online icon |

A Comparison Between Custom Activation Function With Existing Activation Function

Author(s):

Patel Swapnil Shaileshbhai , Vishwakarma Government Engineering College, Chandkheda, Ahmedabad, India

Keywords:

Custom Activation Functions, Existing Activation Functions, Convolutional Neural Networks, CNN, Deep Learning, Performance Comparison, Accuracy, Convergence Speed, Relu, Deep Learning, Computer Vision Tasks, Dataset, Model Architecture, Supervised Learning

Abstract

This research paper investigates the performance and comparison of custom activation functions with existing activation functions in the context of convolutional neural network (CNN) models. The study explores the impact of different activation functions on various datasets and CNN architectures. In particular paper, seven custom activation functions and five existing activation functions are examined to evaluate their effectiveness in enhancing the learning capabilities of CNN models. According to this paper different dataset have different activation function give different accuracy or some custom activation function give good accuracy compare to existing function.

Other Details

Paper ID: IJSRDV11I40145
Published in: Volume : 11, Issue : 4
Publication Date: 01/07/2023
Page(s): 100-106

Article Preview

Download Article