Deep Learning For Noisy Communication System

Abstract

In modern communication systems, channels vary with time due to the mobility of the transmitter (Tx) and the receiver (Rx). To achieve error-free communication in changing environments, Tx and Rx must effectively adapt to different noise levels. However, exiting frameworks face challenges during adaptation, leading to poor communication. In this paper, an independent pre-training collaborative learning framework is designed for tuning Tx and the Rx. The proposed framework incorporates more realistic challenges encountered in communication systems such as lack of feedback channels and the lack of updates at the Rx side. The experimental results show that the proposed approach can reduce noise up to 50% more than existing approaches and within a short training time.