1: \begin{abstract}

2: In this study, we propose an \ac{OAC} scheme for \ac{FEEL}. % without using \ac{CSI} at the \acp{ED} or the \ac{ES}.

3:  The proposed scheme relies on the concept of distributed learning by  \ac{MV} with \ac{signSGD}. As compared to the state-of-the-art solutions, with the proposed method, \acp{ED} transmit the signs of local stochastic gradients by activating one of two orthogonal resources, i.e., \ac{OFDM} subcarriers, and the \acp{MV} at the \ac{ES} are obtained with non-coherent detectors by exploiting the energy accumulations on the subcarriers. Hence, the proposed scheme eliminates the need for \ac{CSI} at the \acp{ED} and \ac{ES}.  By taking path loss, power control, cell size, and the probabilistic nature of the detected \acp{MV} in fading channel  into account, we prove the convergence of the distributed learning for a non-convex function.

4: Through simulations, we show that the proposed scheme can provide  a high test accuracy in fading channels even when the time-synchronization and the power alignment at the \ac{ES} are not ideal. We also provide insight into distributed learning for  location-dependent data distribution for the \ac{MV}-based schemes.

5: 

6: 

7: 

8: \end{abstract}

9: