Time-Triggered Federated Learning Over Wireless Networks

Zhou, Xiaokang; Deng, Yansha; Xia, Huiyun; Wu, Shaochuan; Bennis, Mehdi

doi:10.1109/twc.2022.3189601

Cited by 16 publications

(10 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The FL server aggregated the local model and the latest global model at a cost of stability and convergence delay compared to Sync-FL. A few improvements of Async-FL are Federated Learning with Asynchronous Tiers (Fed-AT) [18] and Time-Triggered Federated Learning (TT-Fed) [19]. Fed-AT and TT-Fed divided an FL process into multiple Sync-FL parallel processes among users with similar processing speeds.…”

Section: Kpi Referencementioning

confidence: 99%

“…In the case of Fed-AT, the global models of the multiple parallel Sync-FL processes were aggregated asynchronously [18]. In the case of TT-Fed, the global models were synchronously aggregated among some parallel Sync-FL processes that were due to aggregate in the same round [19]. In [20], retransmissions of local models were enabled in case of transmission collisions.…”

Section: Kpi Referencementioning

confidence: 99%

“…According to (19), the transmit power of the i-th user assigned to the n-th subchannel, p n,i , can be rewritten using…”

Section: A Variable Substitutionmentioning

confidence: 99%

“…where the newly added constraint (19) preserves the mapping between p n and A n during the variable substitution, and (25…”

Section: P4 : Minmentioning

confidence: 99%

“…(14b), (14c), where (34b) provides the lower bound of A |Kn|−1 (n) and obtained by substituting (19) into…”

Section: Proposition 1 the Following Inequality Always Holdsmentioning

confidence: 99%

See 4 more Smart Citations

Multi-Carrier NOMA-Empowered Wireless Federated Learning with Optimal Power and Bandwidth Allocation

Li¹,

Lv²,

Cao³

et al. 2023

Preprint

View full text Add to dashboard Cite

Wireless federated learning (WFL) undergoes a communication bottleneck in uplink, limiting the number of users that can upload their local models in each global aggregation round. This paper presents a new multi-carrier non-orthogonal multiple-access (MC-NOMA)-empowered WFL system under an adaptive learning setting of Flexible Aggregation. Since a WFL round accommodates both local model training and uploading for each user, the use of Flexible Aggregation allows the users to train different numbers of iterations per round, adapting to their channel conditions and computing resources. The key idea is to use MC-NOMA to concurrently upload the local models of the users, thereby extending the local model training times of the users and increasing participating users. A new metric, namely, Weighted Global Proportion of Trained Mini-batches (WGPTM), is analytically established to measurethe convergence of the new system. Another important aspect is that we maximize the WGPTM to harness the convergence of the new system by jointly optimizing the transmit powers and subchannel bandwidths. This nonconvex problem is converted equivalently to a tractable convex problem and solved efficiently using variable substitution and Cauchy's inequality. As corroborated experimentally using a convolutional neural network and an 18-layer residential network, the proposed MC-NOMA WFL can efficiently reduce communication delay, increase local model training times, and accelerate the convergence by over 40%, compared to its existing alternative.

show abstract

Section: Kpi Referencementioning

confidence: 99%