We present the implementation of binary and ternary neural networks in the hls4ml library, designed to automatically convert deep neural network models to digital circuits with FPGA firmware. Starting from benchmark models trained with floating point precision, we investigate different strategies to reduce the network's resource consumption by reducing the numerical precision of the network parameters to binary or ternary. We discuss the trade-off between model accuracy and resource consumption. In addition, we show how to balance between latency and accuracy by retaining full precision on a selected subset of network components. As an example, we consider two multiclass classification tasks: handwritten digit recognition with the MNIST data set and jet identification with simulated proton-proton collisions at the CERN Large Hadron Collider. The binary and ternary implementation has similar performance to the higher precision implementation while using drastically fewer FPGA resources.
@article{arxiv.2003.06308,
title = {Compressing deep neural networks on FPGAs to binary and ternary precision with HLS4ML},
author = {Giuseppe Di Guglielmo and Javier Duarte and Philip Harris and Duc Hoang and Sergo Jindariani and Edward Kreinar and Mia Liu and Vladimir Loncar and Jennifer Ngadiuba and Kevin Pedro and Maurizio Pierini and Dylan Rankin and Sheila Sagear and Sioni Summers and Nhan Tran and Zhenbin Wu},
journal= {arXiv preprint arXiv:2003.06308},
year = {2020}
}