Federated Learning With Non-IID Data in Wireless Networks

Zhao, Zhongyuan; Feng, Chenyuan; Hong, Wei; Jiang, Jiamo; Jia, Chao; Quek, Tony Q. S.; Peng, Mugen

doi:10.1109/twc.2021.3108197

Cited by 73 publications

(27 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…4) Bayesian Sampler (Bayesian): Search for a sub-optimal L sel devices using Bayesian optimization [47] to meet Eqs. ( 10)- (13). By default, we set the number of initial points to 5 and exploration iterations to 25.…”

Section: A Experiments Setupmentioning

confidence: 99%

“…5) Genetic Sampler (GA): Search for a sub-optimal L sel devices using genetic algorithm [48] to meet Eqs. ( 10)- (13), in which the constrained 0-1 vector solutions are regarded as genes and suffer from selection, crossover, mutation and elimination. By default, we set the population size to 100, the mutation probability to 0.001, and the number of generations to 100.…”

Section: A Experiments Setupmentioning

confidence: 99%

“…Industries (e.g., manufacturing, logistics, and transportation) often face even more serious data threats due to owning a vast amount of valuable information, so they possess the most urgent and critical requirement to increase security to protect data. Therefore, any form of disclosure [10]- [13] or tampering [14]- [16] of the confidential raw data is not allowed. 2) Rapidly changing streaming data on data-intensive sensors.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Data Heterogeneity-Robust Federated Learning via Group Client Selection in Industrial IoT

Li¹,

He²,

Yu³

et al. 2022

Preprint

View full text Add to dashboard Cite

Nowadays, the industrial Internet of Things (IIoT) has played an integral role in Industry 4.0 and produced massive amounts of data for industrial intelligence. These data locate on decentralized devices in modern factories. To protect the confidentiality of industrial data, federated learning (FL) was introduced to collaboratively train shared machine learning models. However, the local data collected by different devices skew in class distribution and degrade industrial FL performance. This challenge has been widely studied at the mobile edge, but they ignored the rapidly changing streaming data and clustering nature of factory devices, and more seriously, they may threaten data security. In this paper, we propose FEDGS, which is a hierarchical cloud-edge-end FL framework for 5G empowered industries, to improve industrial FL performance on non-i.i.d. data. Taking advantage of naturally clustered factory devices, FEDGS uses a gradient-based binary permutation algorithm (GBP-CS) to select a subset of devices within each factory and build homogeneous super nodes participating in FL training. Then, we propose a compound-step synchronization protocol to coordinate the training process within and among these super nodes, which shows great robustness against data heterogeneity. The proposed methods are time-efficient and can adapt to dynamic environments, without exposing confidential industrial data in risky manipulation. We prove that FEDGS has better convergence performance than FedAvg and give a relaxed condition under which FEDGS is more communication-efficient. Extensive experiments show that FEDGS improves accuracy by 3.5% and reduces training rounds by 59% on average, confirming its superior effectiveness and efficiency on non-i.i.d. data.

show abstract

Section: A Experiments Setupmentioning

confidence: 99%

Section: A Experiments Setupmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Data Heterogeneity-Robust Federated Learning via Group Client Selection in Industrial IoT

Li¹,

He²,

Yu³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…This statistical heterogeneity has an impact on the convergence mechanism of FL and lowers model accuracy. One can use an adaptive averaging strategy or apply data sharing mechanism to reduce the impact of non-i.i.d datasets [11]. To cope with the heterogeneity of systems, FL must handle a variety of devices that have differing amounts of memory and processing power as well as differing battery sizes and storage capacities [12].…”

Section: Introductionmentioning

confidence: 99%

Rate-Convergence Tradeoff of Federated Learning over Wireless Channel

Salari¹,

Shirvanimoghaddam²,

Vucetic³

et al. 2022

Preprint

View full text Add to dashboard Cite

In this paper, we consider a federated learning problem over wireless channel that takes into account the coding rate and packet transmission errors. Communication channels are modelled as packet erasure channels (PEC), where the erasure probability is determined by the block length, code rate, and signal-to-noise ratio (SNR). To lessen the effect of packet erasure on the FL performance, we propose two schemes in which the central node (CN) reuses either the past local updates or the previous global parameters in case of packet erasure. We investigate the impact of coding rate on the convergence of federated learning (FL) for both short packet and long packet communications considering erroneous transmissions. Our simulation results shows that even one unit of memory has considerable impact on the performance of FL in erroneous communication.

show abstract

“…For example, in [ 17 ], the client selection and bandwidth allocation for wireless federated learning networks are discussed, and the authors concentrate on the long-term perspective of resource allocation. An interesting analysis of non-independent and identically distributed (non-IID) data processed in dynamically changing wireless networks is presented in [ 18 ], where the averaging scheme is also proposed to reduce the distribution divergence of such kinds of data. Next, in [ 19 ], the energy efficiency is discussed in the context of federated learning over wireless networks.…”

Section: Introductionmentioning

confidence: 99%

Federated Learning for 5G Radio Spectrum Sensing

Wasilewska

Bogucka

Kliks

2021

Sensors

View full text Add to dashboard Cite

Spectrum sensing (SS) is an important tool in finding new opportunities for spectrum sharing. The users, called Secondary Users (SU), who do not have a license to transmit without hindrance, need to employ SS in order to detect and use the spectrum without interfering with the licensed users’ (primary users’ (PUs’)) transmission. Deep learning (DL) has proven to be a good choice as an intelligent SS algorithm that considers radio environmental factors in the decision-making process. It is impossible though for SU to collect the required data and train complex DL models. In this paper, we propose to employ a Federated Learning (FL) algorithm in order to distribute data collection and model training processes over many devices. The proposed method categorizes FL devices into groups by their mean Signal-to-Noise ratio (SNR) and creates a common DL model for each group in the iterative process. The results show that detection accuracy obtained via the FL algorithm is similar to detection accuracy obtained by employing several DL models, namely convolutional neural networks (CNNs), specialized in spectrum detection for a PU signal with a given mean SNR value. At the same time, the main goal of simplification of the SS process in the network is achieved.

show abstract

Federated Learning With Non-IID Data in Wireless Networks

Cited by 73 publications

References 40 publications

Data Heterogeneity-Robust Federated Learning via Group Client Selection in Industrial IoT

Data Heterogeneity-Robust Federated Learning via Group Client Selection in Industrial IoT

Rate-Convergence Tradeoff of Federated Learning over Wireless Channel

Federated Learning for 5G Radio Spectrum Sensing

Contact Info

Product

Resources

About