Meta-learning self-supervised detection based on meta-regularization

Deep learning excels in computer vision and natural language processing, seamlessly intertwining with intelligent communications. Yet, within dynamic channel landscapes, current deep learning models struggle to swiftly acclimate. Thus, a novel meta-learning-infused symbol detection framework is introduced, swiftly adjusting to diverse channel environments. Additionally, a meta-regularization approach addresses data incongruity between support and query sets. The experimental results show that the symbol error rate(SER) is reduced by 37%, which is better than other detection algorithms.