Federated Learning Over Wireless Networks: Convergence Analysis and Resource Allocation

Dinh, Canh T.; Tran, Nguyen H.; Nguyen, Minh N. H.; Hong, Choong Seon; Bao, Wei; Zomaya, Albert Y.; Gramoli, Vincent

doi:10.1109/tnet.2020.3035770

Cited by 325 publications

(197 citation statements)

References 17 publications

Supporting

Mentioning

197

Contrasting

Order By: Relevance

“…There are two challenges for solving the optimization problem in (7). On one hand, due to the limited bandwidth, only a small portion of UEs' training loss and model parameters can be updated to AP.…”

Section: Sliding Differential Evolution Based Schedulingmentioning

confidence: 99%

“…This makes it impossible to calculate the global loss F (w t ) in (7a) accurately. Also, it is difficult to get the relationship between S[t] and w t in (7), where w t relies on model training. On the other hand, this optimization problem is a combinatorial problem which does not normally have low complexity solutions.…”

Section: Sliding Differential Evolution Based Schedulingmentioning

confidence: 99%

“…Additionally, other researchers, e.g. [7], tried to accelerate the convergence by converting the optimization problem into sub-problems. Another challenge is that the limited capacity of communications, e.g., limited bandwidth.…”

Section: Introductionmentioning

confidence: 99%

“…However, there is little literature related to joint energy efficiency and model convergence for federated learning in wireless communication. The authors in [7] carefully analysed the trade-off between energy consumed by UEs and FL convergence with no bandwidth-limited constrains. The authors in [9] proposed an energy-efficient bandwidth resource scheduling policy for FL in wireless networks, and it only considered the communication energy cost.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Sliding Differential Evolution Scheduling for Federated Learning in Bandwidth-Limited Networks

Luo

et al. 2021

IEEE Commun. Lett.

View full text Add to dashboard Cite

Federated learning (FL) in a bandwidth-limited network with energy-limited user equipments (UEs) is underexplored. In this paper, to jointly save energy consumed by the battery-limited UEs and accelerate the convergence of the global model in FL for the bandwidth-limited network, we propose the sliding differential evolution-based scheduling (SDES) policy. To this end, we first formulate an optimization that aims to minimize a weighted sum of energy consumption and model training convergence. Then, we apply the SDES with parallel differential evolution (DE) operations in several small-scale windows, to address the above proposed problem effectively. Compared with existing scheduling policies, the proposed SDES performs well in reducing energy consumption and the model convergence with lower computational complexity.

show abstract

Section: Sliding Differential Evolution Based Schedulingmentioning

confidence: 99%

Section: Sliding Differential Evolution Based Schedulingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Sliding Differential Evolution Scheduling for Federated Learning in Bandwidth-Limited Networks

Luo

et al. 2021

IEEE Commun. Lett.

View full text Add to dashboard Cite

show abstract

“…However, [32] did not consider variability of compute and communication capabilities across clients. Reference [33] proposed FEDL algorithm for allocating radio resources to clients for reducing convergence time. However, we consider the MEC setting with personalized devices where the compute and communication resources of the clients cannot be tuned.…”

Section: Introductionmentioning

confidence: 99%

Coded Computing for Low-Latency Federated Learning Over Wireless Edge Networks

Prakash

Dhakal

Akdeniz

et al. 2021

IEEE J. Select. Areas Commun.

View full text Add to dashboard Cite

Federated learning enables training a global model from data located at the client nodes, without data sharing and moving client data to a centralized server. Performance of federated learning in a multi-access edge computing (MEC) network suffers from slow convergence due to heterogeneity and stochastic fluctuations in compute power and communication link qualities across clients. We propose a novel coded computing framework, CodedFedL, that injects structured coding redundancy into federated learning for mitigating stragglers and speeding up the training procedure. CodedFedL enables coded computing for non-linear federated learning by efficiently exploiting distributed kernel embedding via random Fourier features that transforms the training task into computationally favourable distributed linear regression. Furthermore, clients generate local parity datasets by coding over their local datasets, while the server combines them to obtain the global parity dataset. Gradient from the global parity dataset compensates for straggling gradients during training, and thereby speeds up convergence. For minimizing the epoch deadline time at the MEC server, we provide a tractable approach for finding the amount of coding redundancy and the number of local data points that a client processes during training, by exploiting the statistical properties of compute as well as communication delays. We also characterize the leakage in data privacy when clients share their local parity datasets with the server. Additionally, we analyze the convergence rate and iteration complexity of CodedFedL under simplifying assumptions, by treating CodedFedL as a stochastic gradient descent algorithm. Finally, for demonstrating gains that CodedFedL can achieve in practice, we conduct numerical experiments using practical network parameters and benchmark datasets, in which CodedFedL speeds up the overall training time by up to 15× in comparison to the benchmark schemes.

show abstract