Quantized deep neural network empowering an IM-DD link running in realtime on a field programmable gate array

Mathieu Chagnon, Jouni Siirtola, Tero Rissa, A. Verma · 2019

We demonstrate a communication system whose receiver DSP is an INT5 quantized 9-layer neural network directly extracting information bits processing 1.08 Tb/s of raw data delivering 12.28 Gb/s at a BER of 4×10-5 running in real-time on an FPGA.

Read the paper · More papers on PaperTik