papers

Publications (19)

cs.LG2025

Role of Mixup in Topological Persistence Based Knowledge Distillation for Wearable Sensor Data

Eun Som Jeon, Hongjun Choi, Matthew P. Buman +1

The analysis of wearable sensor data has enabled many successes in several applications. To represent the high-sampling rate time-series with sufficient detail, the use of topologi…

cs.IR2025

VERIRAG: A Post-Retrieval Auditing of Scientific Study Summaries

Shubham Mohole, Hongjun Choi, Shusen Liu +6

Can democratized information gatekeepers and community note writers effectively decide what scientific information to amplify? Lacking domain expertise, such gatekeepers rely on au…

cs.LG2021

Interpretable COVID-19 Chest X-Ray Classification via Orthogonality Constraint

Ella Y. Wang, Anirudh Som, Ankita Shukla +2

Deep neural networks have increasingly been used as an auxiliary tool in healthcare applications, due to their ability to improve performance of several diagnosis tasks. However, t…

cs.CV2020

AMC-Loss: Angular Margin Contrastive Loss for Improved Explainability in Image Classification

Hongjun Choi, Anirudh Som, Pavan Turaga

Deep-learning architectures for classification problems involve the cross-entropy loss sometimes assisted with auxiliary loss functions like center loss, contrastive loss and tripl…

cs.DC2020

Automatic Cross-Replica Sharding of Weight Update in Data-Parallel Training

Yuanzhong Xu, HyoukJoong Lee, Dehao Chen +3

In data-parallel synchronous training of deep neural networks, different devices (replicas) run the same program with different partitions of the training batch, but weight update…

eess.SP2024

Topological Persistence Guided Knowledge Distillation for Wearable Sensor Data

Eun Som Jeon, Hongjun Choi, Ankita Shukla +4

Deep learning methods have achieved a lot of success in various applications involving converting wearable sensor data to actionable health insights. A common application areas is…

cs.CV2024

Fusion is Not Enough: Single Modal Attacks on Fusion Models for 3D Object Detection

Zhiyuan Cheng, Hongjun Choi, James Liang +5

Multi-sensor fusion (MSF) is widely used in autonomous vehicles (AVs) for perception, particularly for 3D object detection with camera and LiDAR sensors. The purpose of fusion is t…

cs.SE2024

ROCAS: Root Cause Analysis of Autonomous Driving Accidents via Cyber-Physical Co-mutation

Shiwei Feng, Yapeng Ye, Qingkai Shi +5

As Autonomous driving systems (ADS) have transformed our daily life, safety of ADS is of growing significance. While various testing approaches have emerged to enhance the ADS reli…

cs.LG2024

Enhancing Accuracy and Parameter-Efficiency of Neural Representations for Network Parameterization

Hongjun Choi, Jayaraman J. Thiagarajan, Ruben Glatt +1

In this work, we investigate the fundamental trade-off regarding accuracy and parameter efficiency in the parameterization of neural network weights using predictor networks. We pr…

cs.RO2021

PHYSFRAME: Type Checking Physical Frames of Reference for Robotic Systems

Sayali Kate, Michael Chinn, Hongjun Choi +2

A robotic system continuously measures its own motions and the external world during operation. Such measurements are with respect to some frame of reference, i.e., a coordinate sy…

cs.CV2022

Understanding the Role of Mixup in Knowledge Distillation: An Empirical Study

Hongjun Choi, Eun Som Jeon, Ankita Shukla +1

Mixup is a popular data augmentation technique based on creating new samples by linear interpolation between two given data samples, to improve both the generalization and robustne…

cs.CV2020

PI-Net: A Deep Learning Approach to Extract Topological Persistence Images

Anirudh Som, Hongjun Choi, Karthikeyan Natesan Ramamurthy +2

Topological features such as persistence diagrams and their functional approximations like persistence images (PIs) have been showing substantial promise for machine learning and c…

cs.CV2025

Intra-class Patch Swap for Self-Distillation

Hongjun Choi, Eun Som Jeon, Ankita Shukla +1

Knowledge distillation (KD) is a valuable technique for compressing large deep learning models into smaller, edge-suitable networks. However, conventional KD frameworks rely on pre…

cs.CV2023

Leveraging Angular Distributions for Improved Knowledge Distillation

Eun Som Jeon, Hongjun Choi, Ankita Shukla +1

Knowledge distillation as a broad class of methods has led to the development of lightweight and memory efficient models, using a pre-trained model with a large capacity (teacher n…

nucl-th2025

Learning nuclear cross sections across the chart of nuclides with graph neural networks

Hongjun Choi, Sinjini Mitra, Jason Brodksy +6

In this work, we explore the use of deep learning techniques to learn how nuclear cross sections change as we add or remove protons and neutrons. As a proof of principle, we focus…

cs.CV2020

Role of Orthogonality Constraints in Improving Properties of Deep Networks for Image Classification

Hongjun Choi, Anirudh Som, Pavan Turaga

Standard deep learning models that employ the categorical cross-entropy loss are known to perform well at image classification tasks. However, many standard models thus obtained of…

cs.CV2026

Oracle-RLAIF: An Improved Fine-Tuning Framework for Multi-modal Video Models using Reinforcement Learning from Ranking Feedback

Derek Shi, Ruben Glatt, Christine Klymko +5

Recent advances in large video-language models (VLMs) rely on extensive fine-tuning techniques that strengthen alignment between textual and visual comprehension. Leading pipelines…

cs.CV2022

Physical Attack on Monocular Depth Estimation with Optimal Adversarial Patches

Zhiyuan Cheng, James Liang, Hongjun Choi +4

Deep learning has substantially boosted the performance of Monocular Depth Estimation (MDE), a critical component in fully vision-based autonomous driving (AD) systems (e.g., Tesla…

cs.CL2025

Personalized Scientific Figure Caption Generation: An Empirical Study on Author-Specific Writing Style Transfer

Jaeyoung Kim, Jongho Lee, Hongjun Choi +1

We study personalized figure caption generation using author profile data from scientific papers. Our experiments demonstrate that rich author profile data, combined with relevant…