Reliable and Fast Recurrent Neural Network Architecture Optimization
Andrés Camero, Jamal Toutouh, Enrique Alba · elib (German Aerospace Center) · 2021
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.