Emulação de um acelerador em hardware para redes neurais utilizando a memória ReRAM

The implementation of new mobile communication systems brings challenges in network structuring, aiming for high data rates, low latency, spectral efficiency, and device density. Neural networks can optimize the circuits, but they face the massive amount of real-time data to be processed. In this study, a hardware accelerator was emulated to speed up the neural network using ReRAM memory. This accelerator was tested on an image identification neural network, resulting in a 1.01% increase in correctly identified images, with an 8x reduction in network computational complexity.