Evolutionary Ensemble Learning for Remote Interference Discrimination in 6G Networks

Atmospheric ducts can result in severe remote interference in time-division duplex (TDD) communication systems of the sixth-generation (6G). The accurate remote interference discrimination is crucial for ensuring communication reliability. In this paper, an evolutionary ensemble learning method is originally proposed, that allows effectively discriminating remote interference in the imbalanced dataset with a low probability of overfitting. In particular, to tackle the problem of reduced classifier generalization arising from the imbalanced dataset, a weighted evolutionary strategy optimized ensemble neural networks (WENNE) is designed for discriminating remote interference. Numerical results illustrate that our proposed method improves the recognition by 20% compared to existing interference discrimination algorithms on the real dataset consisting of a total of 5,520,000 data pieces.