Search papers, labs, and topics across Lattice.
This paper introduces SB-ECC, a score-based decoder that reframes error-correcting code decoding as a continuous-time denoising process guided by parity constraints. A neural denoiser, trained across noise levels, iteratively refines noisy channel observations into valid codewords via a probability-flow ODE. SB-ECC achieves state-of-the-art bit error rates (BER) across a wide range of code/SNR settings, demonstrating an average SNR gain of 0.17dB over existing methods, while also enabling a direct latency-accuracy trade-off via ODE solver selection.
Ditch complex SNR estimation: this new score-based decoder learns to correct errors across noise levels, outperforming existing methods by 0.17dB on average.
Error-correcting codes enable reliable communication, yet practical soft decoding remains challenging across code families and block lengths. We propose SB-ECC, a score-based decoder that casts decoding as continuous-time denoising. A neural denoiser defines a probability-flow ordinary differential equation (ODE) that iteratively updates the noisy channel observation toward a valid codeword, guided by parity constraints. The model is trained across noise levels without time/SNR conditioning, enabling inference without SNR estimation and supporting a direct latency accuracy trade off controlled by the ODE solver budget. We use the raw signed channel observation as input for learning a continuous denoising field. Across 42 code/SNR settings, SB-ECC achieves the best BER in 39/42 entries, with an average SNR gain of 0.17dB and a maximum gain of 0.46dB over the strongest competing baseline, we showed that swapping the solver from Euler to DPM preserves -ln(BER) while reducing end-to-end decoding time by 8.86% on average (up to 12.82%).
