Abstract: This article presents a reconfigurable accelerator for REcurrent Neural networks with fine-grained cOlumn-Wise matrix–vector multiplicatioN (RENOWN). We propose a novel latency-hiding ...