Lookup Table and Neural Network Based Decoders for Real Time Quantum Error Correction

Minimum weight perfect matching (MWPM) is the most popular decoder in surface code for quantum error correction. However, direct implementation of MWPM inference on hardware is not suitable for real-time deployment due its long latency. Thus, lookup-table (LUT) approach has been proposed to resolve this by pre-computing the decoding results. This paper reports a hardware demo of a LUT-based MWPM on FPGA, achieving decoding latency of only 4 clock cycles. Finally, we present the feasibility study of a neural network based solution to address the memory requirement issue of the LUT based method.