ECabc: A feature tuning program focused on Artificial Neural Network hyperparameters
Sanskriti Sharma, Hernan Gelaf-Romer, Travis J. Kessler, John Hunter Mack · The Journal of Open Source Software · 2019
A rich body of literature exists regarding the optimization of artificial neural networks (ANN), a method comprised of densely interconnected adaptive processing units (Hassoun, 2009).This need for optimization arises from a large number of user-set parameters that greatly affect the quality of the ANN's training.Traditionally, this has been done manually.However, presently with higher computing capacity, it is possible to utilize algorithmic approaches which greatly increase the speed of optimization while enhancing model performance.Commonly used optimization routines include genetic algorithms (Castillo, Merelo, Prieto, Rivas, & Romero, 2000) and particle swarm optimizations (Cam, Yetis, & Yildirim, 2015).However, Cam et al. (2015) showed that an Artificial Bee Colony (ABC) performs best by providing higher accuracy and a shorter run time.