arXiv subjects

Kwangsu Kim

Publications and source records attributed to Kwangsu Kim.

9 recordsLinked to original sources

arXiv2026

One-Step Generalization Ratio Guided Optimization for Domain Generalization

Domain Generalization (DG) aims to train models that generalize to unseen target domains but often overfit to domain-specific features, known as undesired correlations. Gradient-based DG methods typically guide gradients in a dominant direction but often inadvertently reinforce spurious correlations. Recent work has employed dropout to regularize overconfident parameters, but has not explicitly adjusted gradient alignment or ensured balanced parameter updates. We propose GENIE (Generalization-ENhancing Iterative Equalizer), a novel optimizer that leverages the One-Step Generalization Ratio (OSGR) to quantify each parameter's contribution to loss reduction and assess gradient alignment. By dynamically equalizing OSGR via a preconditioning factor, GENIE prevents a small subset of parameters from dominating optimization, thereby promoting domain-invariant feature learning. Theoretically, GENIE balances convergence contribution and gradient alignment among parameters, achieving higher OSGR while retaining SGD's convergence rate. Empirically, it outperforms existing optimizers and enhances performance when integrated with various DG and single-DG methods.

cs.LG↗

arXiv2026

Federated Learning for Feature Generalization with Convex Constraints

Federated learning (FL) often struggles with generalization due to heterogeneous client data. Local models are prone to overfitting their local data distributions, and even transferable features can be distorted during aggregation. To address these challenges, we propose FedCONST, an approach that adaptively modulates update magnitudes based on the parameter strength of the global model. This prevents over-emphasizing well-learned parameters while reinforcing underdeveloped ones. Specifically, FedCONST employs linear convex constraints to ensure training stability and preserve locally learned generalization capabilities during aggregation. A Gradient Signal to Noise Ratio (GSNR) analysis further validates the effectiveness of FedCONST in enhancing feature transferability and robustness. As a result, FedCONST effectively aligns local and global objectives, mitigating overfitting and promoting stronger generalization across diverse FL environments, achieving state-of-the-art performance.

cs.LG↗

arXiv2025

COSMOS: Coherent Supergaussian Modeling with Spatial Priors for Sparse-View 3D Splatting

3D Gaussian Splatting (3DGS) has recently emerged as a promising approach for 3D reconstruction, providing explicit, point-based representations and enabling high-quality real time rendering. However, when trained with sparse input views, 3DGS suffers from overfitting and structural degradation, leading to poor generalization on novel views. This limitation arises from its optimization relying solely on photometric loss without incorporating any 3D structure priors. To address this issue, we propose Coherent supergaussian Modeling with Spatial Priors (COSMOS). Inspired by the concept of superpoints from 3D segmentation, COSMOS introduces 3D structure priors by newly defining supergaussian groupings of Gaussians based on local geometric cues and appearance features. To this end, COSMOS applies inter group global self-attention across supergaussian groups and sparse local attention among individual Gaussians, enabling the integration of global and local spatial information. These structure-aware features are then used for predicting Gaussian attributes, facilitating more consistent 3D reconstructions. Furthermore, by leveraging supergaussian-based grouping, COSMOS enforces an intra-group positional regularization to maintain structural coherence and suppress floaters, thereby enhancing training stability under sparse-view conditions. Our experiments on Blender and DTU show that COSMOS surpasses state-of-the-art methods in sparse-view settings without any external depth supervision.

eess.IV↗

arXiv2024

CSTA: CNN-based Spatiotemporal Attention for Video Summarization

Video summarization aims to generate a concise representation of a video, capturing its essential content and key moments while reducing its overall length. Although several methods employ attention mechanisms to handle long-term dependencies, they often fail to capture the visual significance inherent in frames. To address this limitation, we propose a CNN-based SpatioTemporal Attention (CSTA) method that stacks each feature of frames from a single video to form image-like frame representations and applies 2D CNN to these frame features. Our methodology relies on CNN to comprehend the inter and intra-frame relations and to find crucial attributes in videos by exploiting its ability to learn absolute positions within images. In contrast to previous work compromising efficiency by designing additional modules to focus on spatial importance, CSTA requires minimal computational overhead as it uses CNN as a sliding window. Extensive experiments on two benchmark datasets (SumMe and TVSum) demonstrate that our proposed approach achieves state-of-the-art performance with fewer MACs compared to previous methods. Codes are available at https://github.com/thswodnjs3/CSTA.

cs.CV↗

arXiv2021

Generalized Facial Manipulation Detection with Edge Region Feature Extraction

This paper presents a generalized and robust face manipulation detection method based on the edge region features appearing in images. Most contemporary face synthesis processes include color awkwardness reduction but damage the natural fingerprint in the edge region. In addition, these color correction processes do not proceed in the non-face background region. We also observe that the synthesis process does not consider the natural properties of the image appearing in the time domain. Considering these observations, we propose a facial forensic framework that utilizes pixel-level color features appearing in the edge region of the whole image. Furthermore, our framework includes a 3D-CNN classification model that interprets the extracted color features spatially and temporally. Unlike other existing studies, we conduct authenticity determination by considering all features extracted from multiple frames within one video. Through extensive experiments, including real-world scenarios to evaluate generalized detection ability, we show that our framework outperforms state-of-the-art facial manipulation detection technologies in terms of accuracy and robustness.

cs.CV↗

arXiv2021

A Daily Tourism Demand Prediction Framework Based on Multi-head Attention CNN: The Case of The Foreign Entrant in South Korea

Developing an accurate tourism forecasting model is essential for making desirable policy decisions for tourism management. Early studies on tourism management focus on discovering external factors related to tourism demand. Recent studies utilize deep learning in demand forecasting along with these external factors. They mainly use recursive neural network models such as LSTM and RNN for their frameworks. However, these models are not suitable for use in forecasting tourism demand. This is because tourism demand is strongly affected by changes in various external factors, and recursive neural network models have limitations in handling these multivariate inputs. We propose a multi-head attention CNN model (MHAC) for addressing these limitations. The MHAC uses 1D-convolutional neural network to analyze temporal patterns and the attention mechanism to reflect correlations between input variables. This model makes it possible to extract spatiotemporal characteristics from time-series data of various variables. We apply our forecasting framework to predict inbound tourist changes in South Korea by considering external factors such as politics, disease, season, and attraction of Korean culture. The performance results of extensive experiments show that our method outperforms other deep-learning-based prediction frameworks in South Korea tourism forecasting.

cs.LG↗

arXiv2021

FedCCEA : A Practical Approach of Client Contribution Evaluation for Federated Learning

Client contribution evaluation, also known as data valuation, is a crucial approach in federated learning(FL) for client selection and incentive allocation. However, due to restrictions of accessibility of raw data, only limited information such as local weights and local data size of each client is open for quantifying the client contribution. Using data size from available information, we introduce an empirical evaluation method called Federated Client Contribution Evaluation through Accuracy Approximation(FedCCEA). This method builds the Accuracy Approximation Model(AAM), which estimates a simulated test accuracy using inputs of sampled data size and extracts the clients' data quality and data size to measure client contribution. FedCCEA strengthens some advantages: (1) enablement of data size selection to the clients, (2) feasible evaluation time regardless of the number of clients, and (3) precise estimation in non-IID settings. We demonstrate the superiority of FedCCEA compared to previous methods through several experiments: client contribution distribution, client removal, and robustness test to partial participation.

cs.LG↗

arXiv2020

Néel-type skyrmions and their current-induced motion in van der Waals ferromagnet-based heterostructures

Since the discovery of ferromagnetic two-dimensional (2D) van der Waals (vdW) crystals, significant interest on such 2D magnets has emerged, inspired by their appealing properties and integration with other 2D family for unique heterostructures. In known 2D magnets, spin-orbit coupling (SOC) stabilizes perpendicular magnetic anisotropy (PMA). Such a strong SOC could also lift the chiral degeneracy, leading to the formation of topological magnetic textures such as skyrmions through the Dzyaloshinskii-Moriya interaction (DMI). Here, we report the experimental observation of Néel-type chiral magnetic skyrmions and their lattice (SkX) formation in a vdW ferromagnet Fe3GeTe2 (FGT). We demonstrate the ability to drive individual skyrmion by short current pulses along a vdW heterostructure, FGT/h-BN, as highly required for any skyrmion-based spintronic device. Using first principle calculations supported by experiments, we unveil the origin of DMI being the interfaces with oxides, which then allows us to engineer vdW heterostructures for desired chiral states. Our finding opens the door to topological spin textures in the 2D vdW magnet and their potential device application.

cond-mat.mtrl-sci↗

arXiv2019

Magnetic skyrmion artificial synapse for neuromorphic computing

Since the experimental discovery of magnetic skyrmions achieved one decade ago, there have been significant efforts to bring the virtual particles into all-electrical fully functional devices, inspired by their fascinating physical and topological properties suitable for future low-power electronics. Here, we experimentally demonstrate such a device: electrically-operating skyrmion-based artificial synaptic device designed for neuromorphic computing. We present that controlled current-induced creation, motion, detection and deletion of skyrmions in ferrimagnetic multilayers can be harnessed in a single device at room temperature to imitate the behaviors of biological synapses. Using simulations, we demonstrate that such skyrmion-based synapses could be used to perform neuromorphic pattern-recognition computing using handwritten recognition data set, reaching to the accuracy of ~89 percents, comparable to the software-based training accuracy of ~94 percents. Chip-level simulation then highlights the potential of skyrmion synapse compared to existing technologies. Our findings experimentally illustrate the basic concepts of skyrmion-based fully functional electronic devices while providing a new building block in the emerging field of spintronics-based bio-inspired computing.

physics.app-ph↗