1 paper · 1 filter
Jonathan Herbst, Michael Pellauer, Sherief Reda
We introduce a high-throughput neural network accelerator that embeds most network layers directly in hardware, minimizing data transfer and memory usage while preserving a degree…