A Hybrid Chaotic Activation Function for Artificial Neural Networks

Open Webpage

Siobhan Reid, Ken Ferens

Published: 27 Jul 2020, Last Modified: 15 May 202422nd International Conference on AI (ICAI’20) and the 4th International Conference on Applied Cognitive Computing (ACC’20)EveryoneCC BY 4.0
Abstract: Chaos theory has become an integral concept in the modeling and analysis of the human brain. Chaos describes the seemingly random, unpredictable, and irregular behavior in a dynamical system that is fully deterministic. We propose a new ANN which incorporates chaos. The chaotic neurons in the ANN use two AFs: the sigmoid function and the logistic map function. The combination of the two functions will be referred to as the hybrid chaotic activation function (HCAF). The objective of the HCAF is to more accurately represent the complex and chaotic behavior of the brain.