arXiv subjects

Juhyun Park

Publications and source records attributed to Juhyun Park.

14 recordsLinked to original sources

arXiv2026

SCOPE: Entanglement Frontier Escape for Source-Free Class Unlearning

Source-free class unlearning erases whole classes using only the forget data, judged at the representation level, where features can leak a class the head no longer predicts. Existing feature-space erasers answer with one fixed projection, yet forget and retain classes share a representation, so deleting one disturbs the other where they overlap. We prove this tension is a frontier. Every fixed projection that deletes pays a retain cost of at least the retain-readout energy along the forget-discriminant subspace, and erasing that subspace alone attains the floor. The leading source-free erasers all instantiate the form it binds, so the frontier limits the whole class. Conditioning the erasure on the input escapes it. Spectral Conditional Projective Erasure (SCOPE) does so with a single gate, suppressing the forget subspace chiefly on inputs its frozen head's weight scores read as a forget class. It is closed form, needs no retain data or gradient training, and costs orders of magnitude less than retraining. Across five object, face, and speaker benchmarks spanning two modalities and both convolutional and transformer backbones, the frontier predicts the measured retain cost. SCOPE leads the source-free erasers on every benchmark and forget-set size, and at the hardest setting it tops every unlearner, trained methods included.

cs.LG↗

arXiv2026

Latency Prediction for LLM Inference on NPU Systems

Deploying Large Language Models (LLMs) requires exploring a large configuration space spanning parallelization strategies, batching techniques, and scheduling policies. Exhaustive measurement across this space is impractical, making latency prediction essential for system optimization. While NPUs have emerged as accelerators designed for LLM inference, no prediction methodology has been established for them. Specifically, applying prior work to LLM inference latency prediction on NPUs faces three challenges: undisclosed microarchitecture of commercial NPUs, unpredictable compiler optimizations, and latency non-linearity induced by bucketing. We present LENS, a latency estimator that predicts NPU inference latency without information on the microarchitecture or compiler, and captures the non-linear latency induced by bucketing. LENS profiles each bucket with two end-to-end (E2E) measurements and composes the results to predict latency for arbitrary input-output length combinations. We validate LENS across NPUs from multiple vendors, several LLMs, and diverse workloads, achieving a mean prediction error of 2.15\%. We further compare LENS against two methodologically related baselines, confirming the validity of its approach.

cs.DC↗

arXiv2026

ShuntServe: Cost-Efficient LLM Serving on Heterogeneous Spot GPU Clusters

As large language model (LLM) services become widely adopted, the cost of GPU resources for serving these models in cloud environments has emerged as a critical concern. Spot instances offer up to 90% cost savings over on-demand instances, but their frequent interruptions and limited availability pose significant challenges for continuous LLM serving. GPU spot instances, in particular, exhibit lower and more volatile availability than CPU-based instances, making homogeneous clusters that depend on a single GPU type vulnerable to correlated failures. Heterogeneous clusters spanning multiple GPU types can address this by leveraging complementary availability patterns across diverse spot pools, yet existing LLM serving systems are designed for homogeneous environments and suffer from load imbalance when deployed on heterogeneous GPUs. This paper presents ShuntServe, a cost-efficient LLM serving system for heterogeneous spot GPU clusters. ShuntServe employs a roofline model-based analytical serving performance estimator and a dynamic programming-based model placement optimizer that jointly determines node configuration, parallelization strategy, and layer assignment to maximize throughput across heterogeneous GPUs. To enhance fault tolerance when using spot instances, ShuntServe combines output-preserving request migration with concurrent initialization via a shared tensor store, minimizing migration downtime by overlapping replacement node preparation with ongoing serving. Evaluation on Llama-3.1-70B and Qwen3-32B with a heterogeneous AWS cluster of L4, A10G, and L40S GPUs shows that ShuntServe achieves 1.42x and 1.35x higher throughput than state-of-the-art baselines and attains 31.9% and 31.2% cost efficiency improvements over on-demand instances for offline and online serving, respectively.

cs.DC↗

arXiv2026

Retain-Neutral Surrogates for Min-Max Unlearning

Machine unlearning seeks to remove the influence of designated training data while preserving performance on the remaining data. Approximate unlearning can be viewed as a local editing problem; in min-max unlearning, the key local object is the surrogate point at which the retain objective is evaluated. When forget and retain gradients are strongly aligned, an unconstrained forget-maximizing perturbation can move to a surrogate point that increases retain loss. We propose Retain-Orthogonal Surrogate Unlearning (ROSU), which constrains the inner surrogate construction by maximizing first-order forget gain subject to zero first-order retain change under a fixed perturbation budget. This yields a closed-form retain-orthogonal perturbation, a lightweight transported outer update, and amplification along the retain-neutral direction. Our analysis establishes (i) a curvature-controlled second-order bound on retain damage, (ii) a positive-alignment regime in which ROSU strictly reduces surrogate retain loss relative to standard min-max perturbations, and (iii) near-equivalence when the two gradients are nearly orthogonal. Across vision and language benchmarks (CIFAR-10/100, Tiny-ImageNet, TOFU, WMDP), the empirical pattern follows this geometry: ROSU gives its clearest gains in high-coupling regimes while remaining competitive elsewhere.

cs.LG↗

arXiv2026

HCF: Hierarchical Cascade Framework for Distributed Multi-Stage Image Compression

Distributed multi-stage image compression -- where visual content traverses multiple processing nodes under varying quality requirements -- poses challenges. Progressive methods enable bitstream truncation but underutilize available compute resources; successive compression repeats costly pixel-domain operations and suffers cumulative quality loss and inefficiency; fixed-parameter models lack post-encoding flexibility. In this work, we developed the Hierarchical Cascade Framework (HCF) that achieves high rate-distortion performance and better computational efficiency through direct latent-space transformations across network nodes in distributed multi-stage image compression systems. Under HCF, we introduced policy-driven quantization control to optimize rate-distortion trade-offs, and established the edge quantization principle through differential entropy analysis. The configuration based on this principle demonstrates up to 0.6dB PSNR gains over other configurations. When comprehensively evaluated on the Kodak, CLIC, and CLIC2020-mobile datasets, HCF outperforms successive-compression methods by up to 5.56% BD-Rate in PSNR on CLIC, while saving up to 97.8% FLOPs, 96.5% GPU memory, and 90.0% execution time. It also outperforms state-of-the-art progressive compression methods by up to 12.64% BD-Rate on Kodak and enables retraining-free cross-quality adaptation with 7.13-10.87% BD-Rate reductions on CLIC2020-mobile.

cs.CV↗

arXiv2025

Functional Periodic ARMA Processes

Periodicity is a common feature of time series. For finite-dimensional data, periodic autoregressive moving average (ARMA) models have been extensively studied. In functional time series analysis, AR models have been extended to incorporate periodicity, but existing approaches remain incomplete and do not cover the ARMA setting. This paper develops a rigorous theoretical framework for functional periodic ARMA (fPARMA) processes in general separable Hilbert spaces. The proposed model class accommodates periodically varying dependence structures. We derive sufficient conditions for periodic stationarity, the existence of finite moments, and weak dependence. Moreover, we study Yule-Walker-type estimators for the fPAR operators and, in a specific setting, estimators for the fPARMA operators, and establish convergence rates under Sobolev-type regularity assumptions.

stat.ME↗

arXiv2025

Shape Analysis of Euclidean Curves under Frenet-Serret Framework

Geometric frameworks for analyzing curves are common in applications as they focus on invariant features and provide visually satisfying solutions to standard problems such as computing invariant distances, averaging curves, or registering curves. We show that for any smooth curve in R^d, d>1, the generalized curvatures associated with the Frenet-Serret equation can be used to define a Riemannian geometry that takes into account all the geometric features of the shape. This geometry is based on a Square Root Curvature Transform that extends the square root-velocity transform for Euclidean curves (in any dimensions) and provides likely geodesics that avoid artefacts encountered by representations using only first-order geometric information. Our analysis is supported by simulated data and is especially relevant for analyzing human motions. We consider trajectories acquired from sign language, and show the interest of considering curvature and also torsion in their analysis, both being physically meaningful.

stat.ME↗

arXiv2025

Domain Generalization for Person Re-identification: A Survey Towards Domain-Agnostic Person Matching

Person Re-identification (ReID) aims to retrieve images of the same individual captured across non-overlapping camera views, making it a critical component of intelligent surveillance systems. Traditional ReID methods assume that the training and test domains share similar characteristics and primarily focus on learning discriminative features within a given domain. However, they often fail to generalize to unseen domains due to domain shifts caused by variations in viewpoint, background, and lighting conditions. To address this issue, Domain-Adaptive ReID (DA-ReID) methods have been proposed. These approaches incorporate unlabeled target domain data during training and improve performance by aligning feature distributions between source and target domains. Domain-Generalizable ReID (DG-ReID) tackles a more realistic and challenging setting by aiming to learn domain-invariant features without relying on any target domain data. Recent methods have explored various strategies to enhance generalization across diverse environments, but the field remains relatively underexplored. In this paper, we present a comprehensive survey of DG-ReID. We first review the architectural components of DG-ReID including the overall setting, commonly used backbone networks and multi-source input configurations. Then, we categorize and analyze domain generalization modules that explicitly aim to learn domain-invariant and identity-discriminative representations. To examine the broader applicability of these techniques, we further conduct a case study on a related task that also involves distribution shifts. Finally, we discuss recent trends, open challenges, and promising directions for future research in DG-ReID. To the best of our knowledge, this is the first systematic survey dedicated to DG-ReID.

cs.CV↗

arXiv2022

Curvature and Torsion estimation of 3D functional data: A geometric approach to build the mean shape under the Frenet Serret framework

The analysis of curves has been routinely dealt with using tools from functional data analysis. However its extension to multi-dimensional curves poses a new challenge due to its inherent geometric features that are difficult to capture with the classical approaches that rely on linear approximations. We develop an alternative characterization of a mean that reflects shape variation of the curves. Based on a geometric representation of the curves through the Frenet-Serret ordinary differential equations, we introduce a new definition of mean curvature and mean torsion, as well as mean shape through the notion of mean vector field. This new formulation of the mean for multi-dimensional curves allows us to integrate the parameters for the shape features into the unified functional data modelling framework. We formulate the estimation problem of the functional parameters in a penalized regression and develop an efficient algorithm. We demonstrate our approach with both simulated data and real data examples.

stat.ME↗

arXiv2020

Sparse Functional Linear Discriminant Analysis

Functional linear discriminant analysis offers a simple yet efficient method for classification, with the possibility of achieving a perfect classification. Several methods are proposed in the literature that mostly address the dimensionality of the problem. On the other hand, there is a growing interest in interpretability of the analysis, which favors a simple and sparse solution. In this work, we propose a new approach that incorporates a type of sparsity that identifies non-zero sub-domains in the functional setting, offering a solution that is easier to interpret without compromising performance. With the need to embed additional constraints in the solution, we reformulate the functional linear discriminant analysis as a regularization problem with an appropriate penalty. Inspired by the success of $\ell_1$-type regularization at inducing zero coefficients for scalar variables, we develop a new regularization method for functional linear discriminant analysis that incorporates an $L^1$-type penalty, $\int |f|$, to induce zero regions. We demonstrate that our formulation has a well defined solution that contains zero regions, achieving a functional sparsity in the sense of domain selection. In addition, the misclassification probability of the regularized solution is shown to converge to the Bayes error if the data are Gaussian. Our method does not presume that the underlying function has zero regions in the domain, but produces a sparse estimator that consistently estimates the true function whether or not the latter is sparse. Numerical comparisons with existing methods demonstrate this property in finite samples with both simulated and real data examples.

stat.ME↗

arXiv2020

A Mortality Model for Multi-populations: A Semi-Parametric Approach

Mortality is different across countries, states and regions. Several empirical research works however reveal that mortality trends exhibit a common pattern and show similar structures across populations. The key element in analyzing mortality rate is a time-varying indicator curve. Our main interest lies in validating the existence of the common trends among these curves, the similar gender differences and their variability in location among the curves at the national level. Motivated by the empirical findings, we make the study of estimating and forecasting mortality rates based on a semi-parametric approach, which is applied to multiple curves with the shape-related nonlinear variation. This approach allows us to capture the common features contained in the curve functions and meanwhile provides the possibility to characterize the nonlinear variation via a few deviation parameters. These parameters carry an instructive summary of the time-varying curve functions and can be further used to make a suggestive forecast analysis for countries with barren data sets. In this research the model is illustrated with mortality rates of Japan and China, and extended to incorporate more countries.

stat.AP↗

arXiv2019

Mean curvature and mean shape for multivariate functional data under Frenet-Serret framework

The analysis of curves has been routinely dealt with using tools from functional data analysis. However its extension to multi-dimensional curves poses a new challenge due to its inherent geometric features that are difficult to capture with the classical approaches that rely on linear approximations. We propose a new framework for functional data as multidimensional curves that allows us to extract geometrical features from noisy data. We define a mean through measuring shape variation of the curves. The notion of shape has been used in functional data analysis somewhat intuitively to find a common pattern in one dimensional curves. As a generalization, we directly utilize a geometric representation of the curves through the Frenet-Serret ordinary differential equations and introduce a new definition of mean curvature and mean shape through the mean ordinary differential equation. We formulate the estimation problem in a penalized regression and develop an efficient algorithm. We demonstrate our approach with both simulated data and a real data example.

stat.ME↗

arXiv2019

Mapping Areas using Computer Vision Algorithms and Drones

The goal of this paper is to implement a system, titled as Drone Map Creator (DMC) using Computer Vision techniques. DMC can process visual information from an HD camera in a drone and automatically create a map by stitching together visual information captured by a drone. The proposed approach employs the Speeded up robust features (SURF) method to detect the key points for each image frame; then the corresponding points between the frames are identified by maximizing the determinant of a Hessian matrix. Finally, two images are stitched together by using the identified points. Our results show that despite some limitations from the external environment, we could have successfully stitched images together along video sequences.

cs.CV↗

arXiv2008

Local additive estimation

Additive models are popular in high--dimensional regression problems because of flexibility in model building and optimality in additive function estimation. Moreover, they do not suffer from the so-called {\it curse of dimensionality} generally arising in nonparametric regression setting. Less known is the model bias incurring from the restriction to the additive class of models. We introduce a new class of estimators that reduces additive model bias and at the same time preserves some stability of the additive estimator. This estimator is shown to partially relieve the dimensionality problem as well. The new estimator is constructed by localizing the assumption of additivity and thus named {\it local additive estimator}. Implementation can be easily made with any standard software for additive regression. For detailed analysis we explicitly use the smooth backfitting estimator by Mammen, Linton and Nielsen (1999).

stat.ME↗