Context-Aware Embedding Masking Based on Reinforcement Learning for Semantic Multiplexing

Citations

WEB OF SCIENCE

0
Citations

SCOPUS

0

초록

In shared-embedding (SE) multiple access, U users share a single space and user-specific masks provide the attentionbased separation that determines the multiplexing capacity. The original SE design uses state-agnostic masks. We instead learn a context-aware masking policy that maps (SNR, U) to a mask matrix through a low-rank actor trained by proximal policy optimization, rewarding per-user cosine similarity and mask orthogonality. On real BERT embeddings of AG News, it improves cosine similarity at every SNR and cuts the trainingtime mask-orthogonality penalty tenfold. Compared with a fixed-orthogonal scheme, the method shows that the gain originates from context-aware adaptation rather than orthogonality, while preserving recovery.

키워드

Signal to noise ratio; Training; Modeling; Seeds (agriculture); Bit error rate; Ranking (statistics); Learning (artificial intelligence); Transceivers; Context-aware services; Optimization; Shared embedding; semantic communications; embedding masking; multiple access; context-aware reinforcement learning; proximal policy optimization
제목
Context-Aware Embedding Masking Based on Reinforcement Learning for Semantic Multiplexing
저자
Lee, Ki-Ho; Choi, Hyun-Ho; Lee, Jung-Ryun
DOI
10.1109/LWC.2026.3718953
발행일
2026
유형
Article
저널명
IEEE Wireless Communications Letters
권
15
페이지
4459 ~ 4463