Model Agnostic Meta-Learning based Adaptive DecoderA Cognitive Radio Approach

Conventional Deep Learning algorithms have had a significant impact on wireless communication, including channel estimation, signal decoding, power optimization, and resource allocation. However, the performance of the Conventional Deep Learning Model is limited when deployed in a dynamic environment due to slow convergence and the requirement for large datasets. To overcome these limitations, a novel Model-Agnostic Meta-Learning (MAML)-based decoding algorithm is developed in this work. The MAML-based algorithm requires fewer training samples and converges faster in new environments. Numerical findings confirm the benefits of employing meta-learning for signal decoding in wireless systems in a real-time scenario, achieving a minimum decoding accuracy of 80% under highly noisy conditions and 100% at moderate SNR. The data required for the proposed model are obtained from a practical setup using the National Instruments Universal Software Radio Peripheral (NI-USRP), and the performance of the proposed model is experimentally verified.