1: \begin{abstract}

2: In this study, we propose an \ac{OAC} scheme for \ac{FEEL} without \ac{CSI} at the \acp{ED} or the \ac{ES}. The proposed scheme relies on  non-coherent communication techniques for achieving distributed training  by \ac{MV}. In this work, the votes, i.e., the signs of the local gradients, from the \acp{ED} are represented with the \ac{PPM} symbols constructed with \ac{DFT-s-OFDM}. By taking the delay spread and time-synchronization errors into account, the \ac{MV} at the \ac{ES} is obtained  with an energy detector.  Hence, the proposed scheme does not require \ac{CSI} at the \acp{ED} and \ac{ES}. We also prove the convergence of the distributed training when the \ac{MV} is obtained with the proposed scheme under fading channel.

3: %The proposed method naturally reduces the \ac{PMEPR} of the signal as it inherits the properties of the \ac{SC} waveform. 

4: Through simulations, we show that the proposed scheme provides a high test accuracy in fading channels while resulting in lower \ac{PMEPR} symbols.

5: %, as well as how resource utilization affects the effectiveness of \ac{DFT-s-OFDM} in comparison to \ac{OBDA} with \ac{QAM}. 

6: \end{abstract}

7: