arXiv subjects

Jun-Pyo Hong

Publications and source records attributed to Jun-Pyo Hong.

5 recordsLinked to original sources

arXiv2025

Distribution-Level AirComp for Wireless Federated Learning under Data Scarcity and Heterogeneity

Federated Learning (FL) has gained considerable attention as a privacy-preserving and localized approach to implementing edge artificial intelligence (AI). However, conventional FL methods face critical challenges in realistic wireless edge networks, where training data is both limited and heterogeneous, often leading to unstable training and poor generalization. To address these challenges, we propose a Bayesian wireless FL framework that captures model uncertainty via posterior distributions and performs distribution-level aggregation, mitigating local overfitting and client drift. However, this formulation increases communication overhead and prevents the direct use of conventional Over-the-Air Computation (AirComp), which is widely used to improve communication efficiency in standard FL. To overcome this, we develop a transmission-compatible reformulation of posterior aggregation that enables distribution-level Bayesian updates to be computed over the air, along with a closed-form distributed transmit power control strategy derived from convergence analysis under practical wireless impairments. Extensive simulations demonstrate that the proposed framework significantly improves test accuracy and calibration performance compared to conventional FL methods, particularly in data-scarce and heterogeneous environments.

eess.SP↗

arXiv2024

Asynchronous Federated Learning Using Outdated Local Updates Over TDMA Channel

In this paper, we consider asynchronous federated learning (FL) over time-division multiple access (TDMA)-based communication networks. Considering TDMA for transmitting local updates can introduce significant delays to conventional synchronous FL, where all devices start local training from a common global model. In the proposed asynchronous FL approach, we partition devices into multiple TDMA groups, enabling simultaneous local computation and communication across different groups. This enhances time efficiency at the expense of staleness of local updates. We derive the relationship between the staleness of local updates and the size of the TDMA group in a training round. Moreover, our convergence analysis shows that although outdated local updates hinder appropriate global model updates, asynchronous FL over the TDMA channel converges even in the presence of data heterogeneity. Notably, the analysis identifies the impact of outdated local updates on convergence rate. Based on observations from our convergence rate, we refine asynchronous FL strategy by introducing an intentional delay in local training. This refinement accelerates the convergence by reducing the staleness of local updates. Our extensive simulation results demonstrate that asynchronous FL with the intentional delay can rapidly reduce global loss by lowering the staleness of local updates in resource-limited wireless communication networks.

cs.DC↗

arXiv2024

Enabling Multicast Transmission for Spatio-Temporally Asynchronous User Requests in Wireless Environments

The surge in wireless devices and data traffic volume necessitates more efficient transmission methods. Multicasting has garnered consistent attention as a means to fulfill the increasing demand for more efficient data transmission methods. Nevertheless, leveraging multicast wireless networks for spatio-temporally asynchronous data requests poses challenges. In this context, this paper introduces a new multicast mechanism called \emph{set-up based merged multicast (SMMC)} to minimize the delivery time of the requested file in wireless networks by considering the uncertainties inherent in wireless channels. The proposed mechanism comprises two phases. The first phase involves gathering asynchronous requests for a file from users experiencing diverse channel conditions. During this phase, packets of the requested file are transmitted individually in unicast mode within a specified set-up time. Following this, the second phase initiates multicast transmission, which sequentially handles the remaining packets of the file in multicast mode. In the proposed mechanism, we optimize the set-up time and transmission rates of both unicast and multicast modes to minimize the expected file delivery time by jointly taking into account the statistical characteristics of wireless channels, users' locations, and file popularity. Additionally, we also delve into a \emph{fine-tuned SMMC} by utilizing posterior information on the multicast group size and further improve the performance. Our performance evaluations reveal that the proposed SMMC outperforms conventional unicast methods, especially with high-demand data.

eess.SY↗

arXiv2021

Rethinking Coarse-to-Fine Approach in Single Image Deblurring

Coarse-to-fine strategies have been extensively used for the architecture design of single image deblurring networks. Conventional methods typically stack sub-networks with multi-scale input images and gradually improve sharpness of images from the bottom sub-network to the top sub-network, yielding inevitably high computational costs. Toward a fast and accurate deblurring network design, we revisit the coarse-to-fine strategy and present a multi-input multi-output U-net (MIMO-UNet). The MIMO-UNet has three distinct features. First, the single encoder of the MIMO-UNet takes multi-scale input images to ease the difficulty of training. Second, the single decoder of the MIMO-UNet outputs multiple deblurred images with different scales to mimic multi-cascaded U-nets using a single U-shaped network. Last, asymmetric feature fusion is introduced to merge multi-scale features in an efficient manner. Extensive experiments on the GoPro and RealBlur datasets demonstrate that the proposed network outperforms the state-of-the-art methods in terms of both accuracy and computational complexity. Source code is available for research purposes at https://github.com/chosj95/MIMO-UNet.

cs.CV↗

arXiv2019

W-Net: Two-stage U-Net with misaligned data for raw-to-RGB mapping

Recent research on learning a mapping between raw Bayer images and RGB images has progressed with the development of deep convolutional neural networks. A challenging data set namely the Zurich Raw-to-RGB data set (ZRR) has been released in the AIM 2019 raw-to-RGB mapping challenge. In ZRR, input raw and target RGB images are captured by two different cameras and thus not perfectly aligned. Moreover, camera metadata such as white balance gains and color correction matrix are not provided, which makes the challenge more difficult. In this paper, we explore an effective network structure and a loss function to address these issues. We exploit a two-stage U-Net architecture and also introduce a loss function that is less variant to alignment and more sensitive to color differences. In addition, we show an ensemble of networks trained with different loss functions can bring a significant performance gain. We demonstrate the superiority of our method by achieving the highest score in terms of both the peak signal-to-noise ratio and the structural similarity and obtaining the second-best mean-opinion-score in the challenge.

eess.IV↗