# Timely Asynchronous Hierarchical Federated Learning: Age of Convergence

arXiv: https://arxiv.org/abs/2306.12400 · HTML: https://mlsanity.com/paper/2306.12400/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 51 references that are in the map.
- First posted on arXiv 21 Jun 2023.

## Builds on (most cited first, 10 of 10)

- [Federated Learning with Non-IID Data](https://mlsanity.com/paper/1806.00582/) (2018)
- [Asynchronous Federated Optimization](https://mlsanity.com/paper/1903.03934/) (2019)
- [FedOBD: Opportunistic Block Dropout for Efficiently Training Large-scale Neural Networks through Federated Learning](https://mlsanity.com/paper/2208.05174/) (2022)
- [Federated Learning with Flexible Control](https://mlsanity.com/paper/2212.08496/) (2022)
- [Communication Efficient Federated Learning over Multiple Access Channels](https://mlsanity.com/paper/2001.08737/) (2020)
- [Communication-Efficient Federated Learning with Dual-Side Low-Rank Compression](https://mlsanity.com/paper/2104.12416/) (2021)
- [Delay Sensitive Hierarchical Federated Learning with Stochastic Local Updates](https://mlsanity.com/paper/2302.04851/) (2023)
- [Asynchronous Hierarchical Federated Learning](https://mlsanity.com/paper/2206.00054/) (2022)
- [Polynomially Coded Regression: Optimal Straggler Mitigation via Data Encoding](https://mlsanity.com/paper/1805.09934/) (2018)
- [Timely Tracking of a Remote Dynamic Source Via Multi-Hop Renewal Updates](https://mlsanity.com/paper/2304.01989/) (2023)

## Abstract (arXiv)

We consider an asynchronous hierarchical federated learning (AHFL) setting with a client-edge-cloud framework. The clients exchange the trained parameters with their corresponding edge servers, which update the locally aggregated model. This model is then transmitted to all the clients in the local cluster. The edge servers communicate to the central cloud server for global model aggregation. The goal of each client is to converge to the global model, while maintaining timeliness of the clients, i.e., having optimum training iteration time. We investigate the convergence criteria for such a system with dense clusters. Our analysis shows that for a system of n clients with fixed average timeliness, the convergence in finite time is probabilistically guaranteed, if the nodes are divided into O(1) number of clusters, that is, if the system is built as a sparse set of edge servers with dense client bases each.