A mapping f: X→Y is called weak sequence-covering if whenever {ya} is a sequence in Y converging to y ∈ Y, there exist a subsequence {ynk} and xk∈f^-1(ynk)(k∈N) ,x∈f^-1 (y) such that xk→x. The main results are: ...A mapping f: X→Y is called weak sequence-covering if whenever {ya} is a sequence in Y converging to y ∈ Y, there exist a subsequence {ynk} and xk∈f^-1(ynk)(k∈N) ,x∈f^-1 (y) such that xk→x. The main results are: (1) Y is a sequential, Frechet, strongly Frechet space iff every weak sepuence-covering mapping onto Y is quotient, pseudo-open, countably bi-quotient respectively, (2) weak sequence-covering mapping preserves cs-network and certain k-(cs-)networks, thus some new mapping theorems on k-(cs-)notworks are proved.展开更多
智能反射表面(Intelligent Reflecting Surface,IRS)能够对入射其上的信号进行一定的相位和幅度的变换,从而达到信号的精确传输,提高信号的覆盖和传输效率。但是这种优势都是在已知精确的信道状态信息(Channel State Information,CSI)...智能反射表面(Intelligent Reflecting Surface,IRS)能够对入射其上的信号进行一定的相位和幅度的变换,从而达到信号的精确传输,提高信号的覆盖和传输效率。但是这种优势都是在已知精确的信道状态信息(Channel State Information,CSI)的前提下才能达到。基于IRS元件的无源性,精确的CSI很难得到。针对此问题使用压缩感知(Compressive Sensing,CS)算法结合深度学习(Deep Learning,DL)的方法来解决。使用共链路结构来优化传统的压缩感知算法,能够在更低的导频开销和信噪比(Signal to Noise Ratio,SNR)下获得更好的归一化均方误差(Normalized Mean Square Error,NMSE)。将信道估计问题看作为去噪问题,把优化后的CS算法所得结果看作含有噪声的CSI,使用多重深层降噪块网络对其进一步去噪,得到更加精确的CSI。实验表明,所提算法较对比算法在相同SNR下有更好的精度。展开更多
文摘A mapping f: X→Y is called weak sequence-covering if whenever {ya} is a sequence in Y converging to y ∈ Y, there exist a subsequence {ynk} and xk∈f^-1(ynk)(k∈N) ,x∈f^-1 (y) such that xk→x. The main results are: (1) Y is a sequential, Frechet, strongly Frechet space iff every weak sepuence-covering mapping onto Y is quotient, pseudo-open, countably bi-quotient respectively, (2) weak sequence-covering mapping preserves cs-network and certain k-(cs-)networks, thus some new mapping theorems on k-(cs-)notworks are proved.
文摘智能反射表面(Intelligent Reflecting Surface,IRS)能够对入射其上的信号进行一定的相位和幅度的变换,从而达到信号的精确传输,提高信号的覆盖和传输效率。但是这种优势都是在已知精确的信道状态信息(Channel State Information,CSI)的前提下才能达到。基于IRS元件的无源性,精确的CSI很难得到。针对此问题使用压缩感知(Compressive Sensing,CS)算法结合深度学习(Deep Learning,DL)的方法来解决。使用共链路结构来优化传统的压缩感知算法,能够在更低的导频开销和信噪比(Signal to Noise Ratio,SNR)下获得更好的归一化均方误差(Normalized Mean Square Error,NMSE)。将信道估计问题看作为去噪问题,把优化后的CS算法所得结果看作含有噪声的CSI,使用多重深层降噪块网络对其进一步去噪,得到更加精确的CSI。实验表明,所提算法较对比算法在相同SNR下有更好的精度。
