This article introduces Random Error Sampling-based Neuroevolution (RESN), a novel automatic method to optimize recurrent neural network architectures. RESN combines an evolutionary algorithm with a training-free evaluation approach. The results show that RESN achieves state-of-the-art error performance while reducing by half the computational time.
@article{arxiv.2106.15295,
title = {Reliable and Fast Recurrent Neural Network Architecture Optimization},
author = {Andrés Camero and Jamal Toutouh and Enrique Alba},
journal= {arXiv preprint arXiv:2106.15295},
year = {2021}
}