… tion approach for on-device federated learning via over-the-air computation. To improve
the performance and the convergence rate for federated learning, we propose a joint device …

… To allow centralized training without data sharing, federated learning (FL) was proposed
in [4] as a method combining distributed training with central aggregation, and is the focus of …

… Federated learning (FL) tackles the aforementioned concerns by collaboratively training
a … As a typical FL algorithm, the federated averaging (FedAvg) proposed in [12] iteratively …

… This work studied the power control optimization problem for the over-the-air federated
learning over fading channels by taking the gradient statistics into account. The optimal power …

… In this section, we present the learning and communication models for over-the-air FL in
a multi-cell wireless network, taking into account the inter-cell interference in both the downlink …

… of a compromised learning performance due to … Air-FEEL. Different from conventional
power control designs (eg, to minimize the individual mean squared error (MSE) of the over-the-air …

… edged as a major challenge in the federated learning literature, and several strategies
have … focus on over-the-air aggregation [6], [7], [17]–[19]. Specifically, a broadband over-theair …

… DSGD over-the-air; that is, we assume that learning takes … as a distributed over-the-air lossy
computation problem. FL … , and an analog over-theair computation scheme are introduced. …

Over-the-air computation (OAC) is a promising technique to realize fast model aggregation
in the uplink of federated edge learning (FEEL). OAC, however, hinges on accurate channel-…

… channel (MAC), over-theair computation (AirComp) can be … spectrum efficiency of the
over-the-air federated learning (AirFL) … conventional cloud learning, federated learning is …