A Novel Multitask Learning Empowered Codebook Design for Downlink SCMA Networks

Sparse code multiple access (SCMA) is a promising code-domain non-orthogonal (NOMA) scheme for the enabling of massive machine-type communication. In SCMA, design good sparse codebooks and efficient multiuser decoding have attracted tremendous research attention in past few years. This letter aims to leverage deep learning jointly downlink SCMA encoder decoder with aid autoencoder. We introduce novel end-to-end based (E2E-SCMA) framework, under which improved low-complexity are obtained. Compared conventional schemes, our numerical results show that proposed E2E-SCMA leads significant improvements terms error rate computational complexity.