Neural Network-Based Demodulation with CAVIA in Massive Sensor Networks
Kuniyasu Yamamoto, Tomotaka Kimura, Jun Cheng · 2023
We propose a neural network-based demodulator for massive sensor networks in a slotted ALOHA scheme over fading channels. We employ a context adaptation via meta-learning (CAVIA) algorithm to train a common shared weight vector and to adaptively learn a device-specific context parameter. Simulations show that demodulation with CAVIA requires fewer pilots for adaptive learning and outperforms conventional learning.