Seunghyun Hwang

Seunghyun Hwang

Auto-generated profile, based on publicly available information

Seunghyun Hwang

Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

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

Representative Publications

View All Publications
1 January 2023 in Korean Journal of Radiology
Verified
Journal Article
Open access
PDF file
SCSuyon Chang
Suyon Chang
KHKyunghwa Han
Kyunghwa Han
YKYonghan Kwon
Yonghan Kwon
LKLina Kim
Lina Kim
SHSeunghyun Hwang
Seunghyun Hwang
BCByoung Wook Choi
Byoung Wook Choi
Chang S, et al. Korean J Radiol. 2023 Apr;24:e34. https://doi.org/10.3348/kjr.2023.0065

24 March 2025 in IEEE Transactions on Components, Packaging and Manufacturing Technology
Verified
Journal Article
Closed access
VNVinicius C. Do Nascimento
Vinicius C. Do Nascimento
SHSeunghyun Hwang
Seunghyun Hwang
MSMichael Joseph Smith
Michael Joseph Smith
TKTejas Kulkarni
Tejas Kulkarni
QQQiang Qiu
Qiang Qiu
DJDan Jiao
Dan Jiao
The floorplan of chiplets in heterogeneously integrated systems in package (SiP) must consider multiphysics (electrical, thermal, and mechanical) performance and meet positional constraints during optimization. This paper sets forth an efficient framework for chiplet floorplanning subject to positional and multiphysics-performance-based constraints. Traditional multi-physics simulations, often impractical in optimization due to high computational cost, are replaced by a high-fidelity and efficient conditional image generative model via image-based machine learning (ML). This model is accurate and capable of performing real-time prediction of multiphysics performance throughout 3D SiPs. Utilizing the image-based ML model for fast performance assessment, we further accelerate the physical design by developing a novel and highly parallelizable dynamic rank-revealing algorithm for solving the underlying constrained optimization problem. We leverage this algorithm to optimize the position of the chiplets subject to multiphysics performance directly without floorplan representation or convexification techniques while meeting a multitude of constraints. The same machine learning model and constraints are also integrated into a state-of-the-art Corner Block List (CBL) floorplan representation under a Simulated Annealing (SA) optimization framework. The accuracy and efficiency of the proposed optimization method are demonstrated in the floorplanning of chiplets on an interposer subject to thermal constraints, and by comparisons against ML-assisted SA-CBL for performing the same task.

15 March 2021 in arXiv
Verified
Preprint
Open access
PDF file
MJMingi Ji
Mingi Ji
SSSeungjae Shin
Seungjae Shin
SHSeunghyun Hwang
Seunghyun Hwang
GPGibeom Park
Gibeom Park
IMIl-Chul Moon
Il-Chul Moon
Knowledge distillation is a method of transferring the knowledge from a pretrained complex teacher model to a student model, so a smaller network can replace a large teacher network at the deployment stage. To reduce the necessity of training a large teacher model, the recent literatures introduced a self-knowledge distillation, which trains a student network progressively to distill its own knowledge without a pretrained teacher network. While Self-knowledge distillation is largely divided into a data augmentation based approach and an auxiliary network based approach, the data augmentation approach looses its local information in the augmentation process, which hinders its applicability to diverse vision tasks, such as semantic segmentation. Moreover, these knowledge distillation approaches do not receive the refined feature maps, which are prevalent in the object detection and semantic segmentation community. This paper proposes a novel self-knowledge distillation method, Feature Refinement via Self-Knowledge Distillation (FRSKD), which utilizes an auxiliary self-teacher network to transfer a refined knowledge for the classifier network. Our proposed method, FRSKD, can utilize both soft label and feature-map distillations for the self-knowledge distillation. Therefore, FRSKD can be applied to classification, and semantic segmentation, which emphasize preserving the local information. We demonstrate the effectiveness of FRSKD by enumerating its performance improvements in diverse tasks and benchmark datasets. The implemented code is available at https://github.com/MingiJi/FRSKD.

25 December 2024 in IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems
Verified
Journal Article
Closed access
SHSeunghyun Hwang
Seunghyun Hwang
MSMichael Joseph Smith
Michael Joseph Smith
VNVinicius C. Do Nascimento
Vinicius C. Do Nascimento
QQQiang Qiu
Qiang Qiu
CKCheng-Kok Koh
Cheng-Kok Koh
DJDan Jiao
Dan Jiao
Thermal optimization plays a crucial role in the design of advanced systems in package. Due to the large number of thermal simulations needed for full design space exploration, reductions in simulation run-time are critical. Here, we propose a data-driven approach to physics simulation by using neural networks (NNs) to cast the temperature solution process into an image-to-image translation problem. We first model the power generation map, conductivity map, and boundary conditions (BCs) into separate channels of an image. We then generate temperature solutions by training a generative adversarial network, composed of a U-Net shaped generator and a discriminator. The resultant NN model can handle diverse thermal simulation scenarios with accuracy. More importantly, our model can handle BCs, power maps, and physical package designs which are unseen during the training. Experiments show that speed wise, it enables near real-time design, providing a $2581\times $ and $9171\times $ speedup over a custom sparse matrix optimized finite element method and ABAQUS, respectively. Comparisons with state-of-the-art methods have demonstrated the accuracy, efficiency, and versatility of the proposed work.