Jun-Hyung Park

Jun-Hyung Park

Auto-generated profile, based on publicly available information

Jun-Hyung Park

Division of Language & AI Hankuk University of Foreign Studies

Research Focus & Metrics

Academic Career Length

21Years (2006-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
30 September 2022 in ACM Transactions on Embedded Computing Systems
Verified
Journal Article
Open access
JPJun-Hyung Park
Jun-Hyung Park
Kang-Min KimKang-Min Kim
Kang-Min Kim
SLSangkeun Lee
Sangkeun Lee
Deep neural networks typically have extensive parameters and computational operations. Pruning and quantization techniques have been widely used to reduce the complexity of deep models. Both techniques can be jointly used for realizing significantly higher compression ratios. However, separate optimization processes and difficulties in choosing the hyperparameters limit the application of both the techniques simultaneously. In this study, we propose a novel compression framework, termed as quantized sparse training, that prunes and quantizes networks jointly in a unified training process. We integrate pruning and quantization into a gradient-based optimization process based on the straight-through estimator. Quantized sparse training enables us to simultaneously train, prune, and quantize a network from scratch. The empirical results validate the superiority of the proposed methodology over the recent state-of-the-art baselines with respect to both the model size and accuracy. Specifically, quantized sparse training achieves a 135 KB model size in the case of VGG16, without any accuracy degradation, which is 40% of the model size feasible based on the state-of-the-art pruning and quantization approach.

10 July 2025 in IEEE Access
Verified
Journal Article
Open access
YJYoun-Gyu Jin
Youn-Gyu Jin
GWGuixin Wu
Guixin Wu
JSJu-Won Seo
Ju-Won Seo
SPSeong-Jin Park
Seong-Jin Park
SHSung-Ho Hur
Sung-Ho Hur
Kang-Min KimKang-Min Kim
Kang-Min Kim
As the number of households raising companion animals increases, there is a corresponding increase in the demand for veterinary medical services. Furthermore, as companion animals age, the prevalence of chronic disease is increasing, leading companion animal owners to seek more help with long-term health management. However, the shortage of veterinarians and the growing complexity of the consultation process have led to a heavier workload for veterinarians, potentially hindering clinical decision-making efficiency. To address these challenges, we propose AI Veterinary Assistance (AVA), a framework leveraging a large language model. AVA automatically extracts symptoms from consultation records, predicts most probable disease using veterinarian certified disease-symptom database and recommend future consultation question. AVA achieved disease prediction accuracies of 91.4%, 93.4%, and 95.9% for Top-3, Top-5, and Top-10, respectively, along with a symptom extraction accuracy of 79.9% on a consultation records dataset constructed from a veterinarian-certified disease-symptom database. Furthermore, on a real-world dataset, AVA achieved disease prediction accuracies of 38.6%, 43.6%, and 51.5% for Top-3, Top-5, and Top-10, respectively. In both datasets, AVA outperformed baseline methods, demonstrating its effectiveness in supporting clinical decision-making. These results suggest that AVA can help reduce veterinarians’ workload while enhancing the efficiency and reliability of the consultation process.

1 September 2006
Verified
Conference Paper
Open access
JPJun-Hyung Park
Jun-Hyung Park
MKMin-Soo Kim
Min-Soo Kim
BNBong-Nam Noh
Bong-Nam Noh
JJJames B D Joshi
James B D Joshi
With the rapidly increasing complexity of computer systems and the sophistication of hacking tools and techniques, there is a crucial need for computer forensic analysis techniques. Very few techniques exist to support forensic analysis of unknown executable files. The existing techniques primarily inspect executable files to detect known signatures or are based on metadata information. A key goal of such forensic investigation is to identify malicious executable files that hackers might have installed in a targeted system. Finding such malware in a compromised system is difficult because it is hard to identify the purpose of the fragments of executable files. In this paper, we present a similarity-based technique that analyzes targeted executable files to identify a malware present in a compromised system. The technique involves assigning a similarity value to the fragments of executable files present in a compromised hard disk against a set of source files. We present some results based on the comparison of assembly instruction sequences of well-known hacking tools with those of various executable files, and suggest various ways to reduce the false positives

1 January 2024
Verified
Conference Paper
Open access
PDF file
JPJun-Hyung Park
Jun-Hyung Park
HPHyuntae Park
Hyuntae Park
YKYeachan Kim
Yeachan Kim
WLWoosang Lim
Woosang Lim
SLSangKeun Lee
SangKeun Lee
Jun-Hyung Park, Hyuntae Park, Yeachan Kim, Woosang Lim, SangKeun Lee. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing: Industry Track. 2024.

1 January 2024
Verified
Conference Paper
Open access
PDF file
JPJun-Hyung Park
Jun-Hyung Park
YKYeachan Kim
Yeachan Kim
MLMingyu Lee
Mingyu Lee
HPHyuntae Park
Hyuntae Park
SLSangKeun Lee
SangKeun Lee
Jun-Hyung Park, Yeachan Kim, Mingyu Lee, Hyuntae Park, SangKeun Lee. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

1 June 2011 in Journal of Institute of Control, Robotics and Systems
Verified
Journal Article
Open access
BJByung-Jin Jung
Byung-Jin Jung
JPJun-Hyung Park
Jun-Hyung Park
TKTaek-Young Kim
Taek-Young Kim
DKDeuk-Young Kim
Deuk-Young Kim
HMHyungpil Moon
Hyungpil Moon