Machine Learning-Aided Efficient Decoding of Reed-Muller Subcodes

Reed-Muller (RM) codes achieve the capacity of general binary-input memoryless symmetric channels and are conjectured to have a comparable performance that random in terms scaling laws. However, such results established assuming maximum-likelihood decoders for code parameters. Also, RM only admit limited sets rates. Efficient as successive cancellation list (SCL) decoder recently-introduced recursive projection-aggregation (RPA) available at finite lengths. In this paper, we focus on subcodes with flexible We first extend RPA decoding algorithm subcodes. To lower complexity our algorithm, referred subRPA, investigate different approaches prune projections. Next, derive soft-decision based version called soft-subRPA, not improves upon subRPA but also enables differentiable algorithm. Building soft-subRPA then provide framework training machine learning (ML) model search good projections minimize error rate. Training ML achieving very close full-projection significantly smaller number show choice matters significantly, ML-aided projection pruning scheme is able find selection, i.e., negligible degradation compared case, given reasonable