Junnan Yang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Junnan Yang has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Guoqiang Mao
School of Transportation, Southeast University, Nanjing, China
3 shared publications - Ming Ding
CSIRO
3 shared publications Ming Ding
2 shared publications- Youjia Chen
College of Physics and Information Engineering, Fuzhou University, Fuzhou, China
2 shared publications - Zihuai Lin
School of Electrical and Information Engineering, The University of Sydney, Sydney, NSW, Australia
2 shared publications - Athanasios Vasilakos
University of Agder (UiA)
1 shared publication
Latest Publications
- Open Access
- 15 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 7103-7117
XAI-FL: An Explainable Artificial Intelligence-Based Framework and Performance Analysis for Federated Learning
Federated learning (FL), Explainable Artificial Intelligence (XAI), Data Heterogeneity, Robust Aggregation
- 15 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 7, 7103-7117
Open AccessXAI-FL: An Explainable Artificial Intelligence-Based Framework and Performance Analysis for Federated Learning
Federated learning (FL), Explainable Artificial Intelligence (XAI), Data Heterogeneity, Robust Aggregation
- 26 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 11 (6) , 6066-6079
FedHelo: Hierarchical Federated Learning With Loss-Based-Heterogeneity in Wireless Networks
Hierarchical federated learning (HFL) in wireless networks significantly saves communication resources due to edge aggregation conducted in edge mobile computing (MEC) servers. Taking into account the spatially correlated characteristics of data in wireless networks, in this paper, we analyze the performance of HFL with hybrid data distributions, i.e. intra-MEC independent and identically distributed (IID) and inter-MEC non-IID data samples. We derive the upper bound of the difference between the achieved loss and the minimum one, which reveals the impacts of data heterogeneity and global aggregation frequency on the performance of HFL. On this basis, we propose an algorithm named FedHelo which optimizes the aggregation weights and edge/global aggregation frequencies under the constraints of training delay and clients' energy consumption. Our experiments i) verify the obtained theoretical results; ii) demonstrate the performance improvement achieved by FedHelo with the optimal aggregation weights and training/aggregation frequencies, especially in the scenario with high data heterogeneity; and iii) show the preference for edge aggregation in the scenario with a tight delay or client's energy constraint.
- 26 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network Science and Engineering
- Vol. 11 (6) , 6066-6079
FedHelo: Hierarchical Federated Learning With Loss-Based-Heterogeneity in Wireless Networks
Hierarchical federated learning (HFL) in wireless networks significantly saves communication resources due to edge aggregation conducted in edge mobile computing (MEC) servers. Taking into account the spatially correlated characteristics of data in wireless networks, in this paper, we analyze the performance of HFL with hybrid data distributions, i.e. intra-MEC independent and identically distributed (IID) and inter-MEC non-IID data samples. We derive the upper bound of the difference between the achieved loss and the minimum one, which reveals the impacts of data heterogeneity and global aggregation frequency on the performance of HFL. On this basis, we propose an algorithm named FedHelo which optimizes the aggregation weights and edge/global aggregation frequencies under the constraints of training delay and clients' energy consumption. Our experiments i) verify the obtained theoretical results; ii) demonstrate the performance improvement achieved by FedHelo with the optimal aggregation weights and training/aggregation frequencies, especially in the scenario with high data heterogeneity; and iii) show the preference for edge aggregation in the scenario with a tight delay or client's energy constraint.
- 16 April 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 23 (15209210) , 772-783
Frequency-Dependent Depth Map Enhancement via Iterative Depth-Guided Affine Transformation and Intensity-Guided Refinement
Recently, deep convolutional neural network shows significant improvement for intensity-guided depth map enhancement. The most networks focus on either increasing depth or easing features propagation via residual learning and dense connection. However, it has not been explicitly considered yet to mitigate the artifacts caused by the differences of the distributions between the depth map and the corresponding color image, e.g., edge misalignment. In this paper, a novel depth-guided affine transformation is used to filter out the unrelated intensity features, which is further used to refine the depth features. Since the quality of initial depth features is low, the depth-guided intensity features filtering and the intensity-guided depth features refinement are iteratively performed, which progressively promotes effects of such tasks. To make full use of the iterations, all the refined depth features are dense connected followed by a 1$\times$1 convolution layer. In addition, to improve the performance in the case of large upsampling factors (e.g., 16$\times$), the depth features are enhanced from coarse to fine. In each frequency-dependent refinement of the depth features, the above iterative subnetwork as well as the residual learning are introduced. The proposed method is tested for the noise-free and noisy cases which compares against 16 state-of-the-art methods. Our experimental results show the improved performances based on the qualitative and quantitative evaluations.
- 16 April 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Multimedia
- Vol. 23 (15209210) , 772-783
Frequency-Dependent Depth Map Enhancement via Iterative Depth-Guided Affine Transformation and Intensity-Guided Refinement
Recently, deep convolutional neural network shows significant improvement for intensity-guided depth map enhancement. The most networks focus on either increasing depth or easing features propagation via residual learning and dense connection. However, it has not been explicitly considered yet to mitigate the artifacts caused by the differences of the distributions between the depth map and the corresponding color image, e.g., edge misalignment. In this paper, a novel depth-guided affine transformation is used to filter out the unrelated intensity features, which is further used to refine the depth features. Since the quality of initial depth features is low, the depth-guided intensity features filtering and the intensity-guided depth features refinement are iteratively performed, which progressively promotes effects of such tasks. To make full use of the iterations, all the refined depth features are dense connected followed by a 1$\times$1 convolution layer. In addition, to improve the performance in the case of large upsampling factors (e.g., 16$\times$), the depth features are enhanced from coarse to fine. In each frequency-dependent refinement of the depth features, the above iterative subnetwork as well as the residual learning are introduced. The proposed method is tested for the noise-free and noisy cases which compares against 16 state-of-the-art methods. Our experimental results show the improved performances based on the qualitative and quantitative evaluations.
- 9 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 5 (4) , 873-885
Interference Management in In-Band D2D Underlaid Cellular Networks
Recently, it has been standardized by the 3rd Generation Partnership Project (3GPP) roessler that device-to-device (D2D) communications should use uplink resources when coexisting with conventional cellular communications. With uplink resource sharing, both cellular and D2D links cause significant co-channel interference. In this paper, we address the critical issue of interference management in the network considering a practical path loss model incorporating both line-of-sight (LoS) and non-line-of-sight (NLoS) transmissions. To reduce the severe interference caused by active D2D links, we consider a mode selection scheme based on the maximum received signal strength (MRSS) for each user equipment (UE) to control the D2D-to-cellular interference. Furthermore, we analyze the performance in terms of the coverage probability and the area spectral efficiency (ASE) for both the cellular network and the D2D one. Analytical results are obtained and the accuracy of the proposed analytical framework is validated through Monte Carlo simulations. Through our theoretical and numerical analyses, we quantify the performance gains brought by D2D communications in cellular networks and we find an optimum mode selection threshold β to maximize the total ASE in the network. This insight is expected to provide a design guideline for D2D mode selections.
- 9 July 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Cognitive Communications and Networking
- Vol. 5 (4) , 873-885
Interference Management in In-Band D2D Underlaid Cellular Networks
Recently, it has been standardized by the 3rd Generation Partnership Project (3GPP) roessler that device-to-device (D2D) communications should use uplink resources when coexisting with conventional cellular communications. With uplink resource sharing, both cellular and D2D links cause significant co-channel interference. In this paper, we address the critical issue of interference management in the network considering a practical path loss model incorporating both line-of-sight (LoS) and non-line-of-sight (NLoS) transmissions. To reduce the severe interference caused by active D2D links, we consider a mode selection scheme based on the maximum received signal strength (MRSS) for each user equipment (UE) to control the D2D-to-cellular interference. Furthermore, we analyze the performance in terms of the coverage probability and the area spectral efficiency (ASE) for both the cellular network and the D2D one. Analytical results are obtained and the accuracy of the proposed analytical framework is validated through Monte Carlo simulations. Through our theoretical and numerical analyses, we quantify the performance gains brought by D2D communications in cellular networks and we find an optimum mode selection threshold β to maximize the total ASE in the network. This insight is expected to provide a design guideline for D2D mode selections.
- Open Access
- 14 March 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 35755-35768
Analysis of Underlaid D2D-Enhanced Cellular Networks: Interference Management and Proportional Fair Scheduler
Device-to-device (D2D) communications have been proposed as a promising technology to improve network capacity and user experiences in the future mobile networks such as heterogeneous networks with densely deployed small cells, but it has not yet been fully incorporated into the existing cellular networks. Interference management is one of the critical issues when D2D communications using uplink resources and coexisting with conventional cellular communications, especially in the ultra-dense networks (UNDs). In this paper, we address the critical issue of interference management by a mode selection method which is based on the maximum received signal strength (MRSS) for each D2D transmitter (TU). To analyze the capacity of a more practical D2D-enhanced network, we consider that the typical user is no longer a random user, i.e., random user selection by a round-robin (RR) scheduler, as assumed in most studies in the literature. Instead, a cellular user with the maximum proportional fair (PF) metric is chosen by its serving BS as the typical user, which is referred to as the PF scheduler in the cellular tier. Furthermore, we theoretically study the performance in terms of the coverage probability and the area spectral efficiency (ASE) for both the cellular network and the D2D one with the consideration of the PF scheduler in UDNs. Analytical results are obtained and the accuracy of the proposed analytical framework is validated through Monte Carlo simulations. Through our theoretical and numerical analyses, we quantify the performance gains brought by D2D communications and the PF scheduler in cellular networks and we find an optimum mode selection threshold β to maximize the total ASE in the network.
- 14 March 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 35755-35768
Open AccessAnalysis of Underlaid D2D-Enhanced Cellular Networks: Interference Management and Proportional Fair Scheduler
Device-to-device (D2D) communications have been proposed as a promising technology to improve network capacity and user experiences in the future mobile networks such as heterogeneous networks with densely deployed small cells, but it has not yet been fully incorporated into the existing cellular networks. Interference management is one of the critical issues when D2D communications using uplink resources and coexisting with conventional cellular communications, especially in the ultra-dense networks (UNDs). In this paper, we address the critical issue of interference management by a mode selection method which is based on the maximum received signal strength (MRSS) for each D2D transmitter (TU). To analyze the capacity of a more practical D2D-enhanced network, we consider that the typical user is no longer a random user, i.e., random user selection by a round-robin (RR) scheduler, as assumed in most studies in the literature. Instead, a cellular user with the maximum proportional fair (PF) metric is chosen by its serving BS as the typical user, which is referred to as the PF scheduler in the cellular tier. Furthermore, we theoretically study the performance in terms of the coverage probability and the area spectral efficiency (ASE) for both the cellular network and the D2D one with the consideration of the PF scheduler in UDNs. Analytical results are obtained and the accuracy of the proposed analytical framework is validated through Monte Carlo simulations. Through our theoretical and numerical analyses, we quantify the performance gains brought by D2D communications and the PF scheduler in cellular networks and we find an optimum mode selection threshold β to maximize the total ASE in the network.
- 11 February 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 18 (3) , 1779-1791
Optimal Base Station Antenna Downtilt in Downlink Cellular Networks
Very recent studies showed that the area spectral efficiency (ASE) of downlink (DL) cellular networks will continuously decrease and finally crash to zero as the base station (BS) density increases towards infinity if the absolute height difference between BS antenna and user equipment (UE) antenna is larger than zero. Such a phenomenon is referred to as the ASE Crash. We revisit this issue by considering optimizing the BS antenna downtilt in cellular networks. It is common to adjust antenna pattern to tune the direction of the vertical beamforming, and thus increasing received signal power and/or reducing inter-cell interference power to improve network performance. This paper focuses on investigating the relationship between the BS antenna downtilt and the downlink network performance in terms of the coverage probability and the ASE. Our results reveal an interesting find that there exists an optimal antenna downtilt to achieve the maximum coverage probability for each BS density. Numerically solvable expressions are derived for such optimal antenna downtilt, which is a function of the BS density. Our numerical results show that after applying the optimal antenna downtilt, the network performance can be significantly improved, and hence the ASE crash can be delayed by nearly one order of magnitude in terms of the BS density. Our results also give guidance on setting the optimum downtilt angle to maximize network performance given a fixed BS density.
- 11 February 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Wireless Communications
- Vol. 18 (3) , 1779-1791
Optimal Base Station Antenna Downtilt in Downlink Cellular Networks
Very recent studies showed that the area spectral efficiency (ASE) of downlink (DL) cellular networks will continuously decrease and finally crash to zero as the base station (BS) density increases towards infinity if the absolute height difference between BS antenna and user equipment (UE) antenna is larger than zero. Such a phenomenon is referred to as the ASE Crash. We revisit this issue by considering optimizing the BS antenna downtilt in cellular networks. It is common to adjust antenna pattern to tune the direction of the vertical beamforming, and thus increasing received signal power and/or reducing inter-cell interference power to improve network performance. This paper focuses on investigating the relationship between the BS antenna downtilt and the downlink network performance in terms of the coverage probability and the ASE. Our results reveal an interesting find that there exists an optimal antenna downtilt to achieve the maximum coverage probability for each BS density. Numerically solvable expressions are derived for such optimal antenna downtilt, which is a function of the BS density. Our numerical results show that after applying the optimal antenna downtilt, the network performance can be significantly improved, and hence the ASE crash can be delayed by nearly one order of magnitude in terms of the BS density. Our results also give guidance on setting the optimum downtilt angle to maximize network performance given a fixed BS density.
- Open Access
Analysis of Device-to-Device Communications in Uplink Cellular Networks with Lognormal Fading
In this paper, using the stochastic geometry theory, we present a framework for analyzing the performance of device-to-device (D2D) communications underlaid uplink (UL) cellular networks. In our analysis, we consider a D2D mode selection criterion based on an energy threshold for each user equipment (UE). Specifically, a UE will operate in a cellular mode, if its received signal strength from the strongest base station (BS) is large than a threshold \beta. Otherwise, it will operate in a D2D mode. Furthermore, we consider a generalized log-normal shadowing in our analysis. The coverage probability and the area spectral efficiency (ASE) are derived for both the cellular network and the D2D one. Through our theoretical and numerical analyses, we quantify the performance gains brought by D2D communications and provide guidelines of selecting the parameters for network operations.
Open AccessAnalysis of Device-to-Device Communications in Uplink Cellular Networks with Lognormal Fading
In this paper, using the stochastic geometry theory, we present a framework for analyzing the performance of device-to-device (D2D) communications underlaid uplink (UL) cellular networks. In our analysis, we consider a D2D mode selection criterion based on an energy threshold for each user equipment (UE). Specifically, a UE will operate in a cellular mode, if its received signal strength from the strongest base station (BS) is large than a threshold \beta. Otherwise, it will operate in a D2D mode. Furthermore, we consider a generalized log-normal shadowing in our analysis. The coverage probability and the area spectral efficiency (ASE) are derived for both the cellular network and the D2D one. Through our theoretical and numerical analyses, we quantify the performance gains brought by D2D communications and provide guidelines of selecting the parameters for network operations.