Publications

Privacy

1. 

   Minimax optimal two-sample testing under local differential privacy

   Jongmin Mun, Seungwoo Kwak, and Ilmun Kim

   arxiv preprint, Nov 2024

   Summary Paper

   We explore the trade-off between privacy and statistical utility in private two-sample testing under local differential privacy (LDP) for both multinomial and continuous data. We begin by addressing the multinomial case, where we introduce private permutation tests using practical privacy mechanisms such as Laplace, discrete Laplace, and Google’s RAPPOR. We then extend our multinomial approach to continuous data via binning and study its uniform separation rates under LDP over H}"older and Besov smoothness classes. The proposed tests for both discrete and continuous cases rigorously control the type I error for any finite sample size, strictly adhere to LDP constraints, and achieve minimax separation rates under LDP. The attained minimax rates reveal inherent privacy-utility trade-offs that are unavoidable in private testing. To address scenarios with unknown smoothness parameters in density testing, we propose an adaptive test based on a Bonferroni-type approach that ensures robust performance without prior knowledge of the smoothness parameters. We validate our theoretical findings with extensive numerical experiments and demonstrate the practical relevance and effectiveness of our proposed methods.

Imbalance

1. 

   Weighted support vector machine for extremely imbalanced data

   Jongmin Mun, Sungwan Bang, and Jaeoh Kim

   Computational Statistics & Data Analysis, Mar 2025

   Summary Paper

   Data augmentation and loss function adjustments are two common techniques for imbalanced classification. In cases of extreme class imbalance, it is often standard practice to combine these two approaches. However, determining the optimal oversampling ratio and the degree of asymmetry in the loss function typically relies on heuristics. We further consider the scenario where the minority class consists of subgroups and propose a straightforward method for combining data augmentation and loss function adjustments. This method serves as a sample-based approximation of the population-level asymptotically Bayes optimal oracle procedure.

2. 

   Sex differences in autism spectrum disorder using class imbalance adjusted functional connectivity

   Jong Young Namgung, Jongmin Mun, Yeongjun Park, Jaeoh Kim, and Bo-yong Park

   NeuroImage, Dec 2024

   Preliminary version presented at the \emph2024 International Conference of the IEEE Engineering in Medicine and Biology Society

   Summary Paper

   Detecting sex differences in brain connectome organization in autism spectrum disorder (ASD) through statistical hypothesis testing is challenging. This challenge arises from the reliance of statistical power on the smaller group size, coupled with the significant imbalance in diagnostic ratios of ASD between males and females. While Chen et al. (2013, Journal of the American Statistical Association) addressed this issue using an ensemble of undersampling approaches, we take an alternative approach: oversampling. Our method combines dimension reduction via diffusion map embedding with Gaussian mixture model-based oversampling to balance the sex ratio in functional connectivity data. This approach reveals significant sex-related differences in the sensorimotor, attention, and default mode networks, with gradients linked to higher-order cognitive control. Transcriptomic analysis identifies gene enrichment in the cortex, thalamus, and striatum. Notably, gradient-symptom associations differ by sex, with stronger effects observed in females. These findings highlight the sex heterogeneity in ASD’s large-scale brain networks.

3. 

   Prediction of forest fire risk for artillery military training using weighted support vector machine for imbalanced data

   Ji Hyun Nam, Jongmin Mun, Seongil Jo, and Jaeoh Kim

   Journal of Classification, Mar 2024

   Summary Paper

   Artillery training inherently poses wildfire risks. Predictive modeling of these wildfires faces two key challenges: the scarcity of wildfire cases and the limited granularity of meteorological data during training. We address the first challenge by augmenting the data using a Gaussian mixture generative model and adjusting the loss function of a support vector machine. To tackle the second challenge, we integrate the Republic of Korea Army (ROKA) dataset with the Korea Meteorological Administration database. Our resulting model achieves a 99% improvement in balanced classification metrics compared to previous models.

Brain

1. 

   In-vivo integration of soft neural probes through high-resolution printing of liquid electronics on the cranium

   Young-Geun Park, Yong Won Kwon, Chin Su Koh, Enji Kim, Dong Ha Lee, Sumin Kim, Jongmin Mun, Yeon-Mi Hong, Sanghoon Lee, Ju-Young Kim, Jae-Hyun Lee, Hyun Ho Jung, Jinwoo Cheon, Jin Woo Chang, and Jang-Ung Park

   Nature Communications, Feb 2024

   Summary Paper

   Conventional neural probes are rigid, causing damages when implanted in freely moving animals. To address this issue, we present a soft, conformable neural interface system designed for long-term, stable monitoring of single-unit neural activity in freely moving subjects. This system integrates soft neural probes implanted in the brain with liquid metal-based electronics printed on the cranial surface. In-vivo mouse studies, employing dimensionality reduction, clustering analysis, and goodness-of-fit testing, provide statistical evidence of stable neural recordings over 33 weeks and behavior-induced activation across multiple brain regions during T-maze tests.