Research

Research

Core Research

Discretized Diffusion Dynamics

Diffusion processes offer a powerful mathematical framework for understanding and designing machine learning algorithms. By developing a non-asymptotic theory for discretized diffusion, I aim to understand the effects of algorithmic parameters obscured in the asymptotic limit. In my works, this framework is applied to understand and improve diffusion models for generative AI, stochastic gradient descent, and reinforcement learning.

Core Theory

Non-asymptotic analysis of stochastic dynamics

Generative AI

Diffusion Models

Optimization & Reinforcement Learning

Stochastic Approximation

Generative AI

Diffusion models generate realistic data using a denoising process. The rate at which noise is injected to or removed from data determines the quality of generated data, and its effect is captured by non-asymptotic theory.

  1. Noise Schedule Design for Diffusion Models: An Optimal Control Perspective

    Preprint 路 2026

    S. T. Kong, W. Wang, R. Srikant

  2. Direct Preference Optimization for LLM Alignment

    Preprint 路 2025

    Y. Du, S. T. Kong, R. Srikant

Finite-Time Analysis of Stochastic Approximation

  1. Finite-Sample Wasserstein Error Bounds and Concentration Inequalities for Nonlinear Stochastic Approximation

    Preprint 路 2026

    S. T. Kong, R. Srikant

  2. Nonasymptotic CLT and Error Bounds for Two-Time-Scale Stochastic Approximation

    Under review 路 IEEE TAC 路 2025

    S. T. Kong, S. Zeng, T. T. Doan, R. Srikant

  3. When to Use Polyak-Ruppert Averaging in Stochastic Approximation

    Work in progress 路 2026

    S. T. Kong, R. Srikant

Additional Research & Publications

General ML

  1. Spectral Clustering for Crowdsourcing with Inherently Distinct Task Types

    TMLR 路 2025

    S. Mandal*, S. T. Kong*, D. Katselis, R. Srikant

  2. Key Feature Replacement of In-Distribution Samples for Out-of-Distribution Detection

    AAAI 路 2023

    J. Kim*, S. T. Kong*, D. Na, K. H. Jung

  3. A Neural Pre-Conditioning Active Learning Algorithm to Reduce Label Complexity

    NeurIPS 路 2022

    S. T. Kong, S. Jeon, D. Na, J. Lee, H. S. Lee, K. H. Jung

Machine Learning for Clinical Decision-Making

Chest radiograph with a radiologist-annotated pulmonary nodule in the left upper lung.
model confidence 0.92
Radiology Report
Findings
  • Solitary subcentimeter nodule in the left upper lung zone. AI 路 0.92
  • No prior imaging available for comparison.
Impression
  • Indeterminate subcentimeter left upper zone nodule.
  • Recommend non-contrast chest CT for further characterization.
Clinical decision-making: a model flags a nodule, and a radiologist confirms and issues the report and recommendation.

  1. Clinical Utility of Deep Learning Assistance for Detecting Various Abnormal Findings in Color Retinal Fundus Images: A Reader Study

    Translational Vision Science & Technology 路 2024

    J. Y. Shin, J. Son, S. T. Kong, J. Park, B. Park, K. H. Park, K. H. Jung, S. J. Park

  2. Self-Accumulative Vision Transformer for Bone Age Assessment Using the Sauvegrain Method

    ECCV Workshop 路 2024

    H. J. Choi, D. Na, K. Cho, B. Bae, S. T. Kong, H. Ahn, S. Choi, J. Kim

  3. Volume Is All You Need: Improving Multi-Task Multiple Instance Learning for WMH Segmentation and Severity Estimation

    MICCAI Workshop (ML in Clinical Neuroimaging) 路 2022

    W. Jung, C. H. Suh, W. H. Shim, J. Kim, D. Lee, C. Park, S. T. Kong, K. H. Jung, H. Heo, S. J. Kim

  4. Augmenting Magnetic Resonance Imaging with Tabular Features for Enhanced and Interpretable Medial Temporal Lobe Atrophy Prediction

    MICCAI Workshop (ML in Clinical Neuroimaging) 路 2022

    D. Lee, C. H. Suh, J. Kim, W. Jung, C. Park, K. H. Jung, S. T. Kong, W. H. Shim, H. Heo, S. J. Kim

  5. Abstraction in Pixel-Wise Noisy Annotations Can Guide Attention to Improve Prostate Cancer Grade Assessment

    MICCAI Workshop (Limited & Noisy Data) 路 2022

    H. Kim, S. T. Kong, H. Lee, K. Kim, K. H. Jung

  6. Utilizing Synthetic Nodules for Improving Nodule Detection in Chest Radiographs

    Journal of Digital Imaging 路 2022

    M. Chung*, S. T. Kong*, et al.

  7. Self-Supervised Learning with Electrocardiogram Delineation for Arrhythmia Detection

    IEEE EMBC 路 2021

    B. T. Lee*, S. T. Kong*, et al.

  8. Aggregation of Cohorts for Histopathological Diagnosis with Deep Morphological Analysis

    Scientific Reports 路 2021

    J. Park, Y. R. Chung, S. T. Kong, Y. W. Kim, H. Park, K. Kim, D. I. Kim, K. H. Jung

  9. Leveraging the Generalization Ability of Deep Convolutional Neural Networks for Improving Classifiers for Color Fundus Photographs

    Applied Sciences 路 2021

    J. Son, J. Kim, S. T. Kong, K. H. Jung

  10. Manifold Ordinal-Mixup for Ordered Classes in TW3-Based Bone Age Assessment

    MICCAI 路 2020

    B. Bae, J. Lee, S. T. Kong, J. Sung, K. H. Jung

Others

  1. Almost Boltzmann Exploration

    arXiv 路 2019

    H. Gupta, S. T. Kong, R. Srikant, W. Wang

  2. Structure Identification in Layered Precedence Networks

    IEEE CCTA 路 2017

    S. T. Kong, D. Katselis, C. L. Beck, R. Srikant

* Equal contribution.

Next

See my full background

Service & talks View CV