ML-Based Low-Density Spreading Code Design

Low-density sparse codes (LDSC) are a promising technology with applications in multiple access, serving as a crucial element in non-orthogonal multiple access (NOMA) and facilitating feature generation in communication systems. However, most of the studies on LDSC do not include how to refine the LDSC design to improve the system performance when applied to multiple access and spreading. In this paper, we explore the enhancement of LDSC design by leveraging feature engineering and machine learning to generate near-optimal LDSC codes. We illustrate the application of this approach using direct-sequence spreading spectrum (DSSS) in the Internet of Things (IoT) as a test case. Our work is poised to make a substantial contribution to next-generation ML-optimized networks by providing valuable insights and design framework for optimizing and refining LDSC designs tailored to specific application needs.