期刊文献+

一种基于互信息神经估计改进的未知信道端到端通信系统

An Improved End-to-End Unknown Channel Communication System Based on Mutual Information Neural Estimation
下载PDF
导出
摘要 针对端到端通信系统在信道未知时通过使用生成对抗网络(Generative Adversarial Network,GAN)表示信道效应而产生的梯度消失和过拟合问题,提出了一种通过互信息作为正则化项的改进方案。通过互信息神经估计器(Mutual Information Neural Estimator,MINE)计算发射信号与接收信号的互信息并由此作为正则化项来提高模型的泛化能力,从而缓解过拟合问题,而其中产生的额外梯度可以缓解梯度消失问题,从而提高了模型的学习性能。从仿真结果来看,改进后的方案表现出了更好的鲁棒性,并且在加性高斯白噪声(Additive White Gaussian Noise,AWGN)信道、瑞利衰落信道上获得了更好的误比特率性能。此外,在频率选择性信道上,改进后的方案获得了1 dB左右误比特率性能的改善。 To solve the problem of gradient disappearance and overfitting caused by the use of generative adversarial network(GAN)to represent channel effects when the channel is unknown in end-to-end communication systems,an improved scheme using mutual information as a regularization term is proposed.The Mutual Information Neural Estimator(MINE)is used to calculate the mutual information of the transmitted signal and the received signal,and the mutual information is used as a regularization term to improve the generalization ability of the model,thus alleviating the overfitting problem,and the extra gradient generated in it can alleviate the gradient disappearance problem.Thus,the learning performance of the model is improved.From the simulation results,the improved scheme shows better robustness and better bit error rate(BER)performance on additive white Gaussian noise(AWGN)channel and Rayleigh fading channel.In addition,on frequency-selective channels,the improved scheme can improve the BER of about 1 dB.
作者 安永丽 方斌 刘劲芸 纪占林 AN Yongli;FANG Bin;LIU Jinyun;JI Zhanlin(College of Artificial Intelligence,North China University of Technology,Tangshan 063000,China;Key Laboratory of Industrial Intelligent Perception of Hebei Province,Tangshan 063000,China)
出处 《电讯技术》 北大核心 2024年第9期1386-1393,共8页 Telecommunication Engineering
基金 国家科技部重点研发专项(2017YFE0135700) 河北省高层次人才工程项目(A201903011) 河北省自然科学基金(F2018209358)。
关键词 端到端学习 生成对抗网络 互信息 神经估计 end-to-end learning generative adversarial network mutual information neural estimation
  • 相关文献

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部