What is it about?
An activation function is part of a deep neural network that plays a crucial role in its performance. Usually a standard function is selected. We use an evolutionary algorithm to design better functions that increase a network's performance, thus attaining better classification of images.
Featured Image
Photo by Eugene Zhyvchik on Unsplash
Why is it important?
A unique aspect of our work is the use of coevolution, wherein multiple populations coevolve in unison.
Perspectives
Writing this article was a pleasure as it involves a cool idea and a talented grad student.
Professor Moshe Sipper
Ben-Gurion University of the Negev
Read the Original
This page is a summary of: Evolution of activation functions for deep learning-based image classification, July 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3520304.3533949.
You can read the full text:
Resources
Contributors
The following have contributed to this page
