arXiv subjects

Junghwan Lee

Publications and source records attributed to Junghwan Lee.

16 recordsLinked to original sources

arXiv2025

Two-point correlators in de Sitter-prepared states with bra-ket wormholes

Motivated by the finiteness of de Sitter (dS) horizon entropy, we study how "bra-ket wormholes" modify correlation functions in gravitationally prepared states. Euclidean wormhole saddles in gravitational path integrals can generate non-factorizing contributions to correlation functions, as in replica-wormhole explanation of the Page curve and bra-ket-wormhole restoration of strong subadditivity. By defining 'time' variables and computing observables in a flat region attached to the dS boundary, we evaluate bra-ket wormhole contributions to scalar two-point functions and find late-time transitions in the dominant saddle, accompanied by the ramp-and-plateau behavior of correlations and the characteristic timescale comparable to the fast scrambling. Each observable is consistent with `complementarity', in the sense that wormhole effects are distinguishable only at late respective times. Consistencies are based upon the interplay of (i) inflationary horizon exit and re-entry, (ii) enhancement of correlations at small comoving momentum by wormhole contributions, (iii) a competition between mode counting and topological suppression that drives a transition to wormhole dominance, which naturally yields the fast scrambling timescale, and (iv) irreducible errors by cosmic variance in early CMB-like observations. To clearly interpret in terms of entropy and chaotic nature of dS, one needs a more complete mechanism of wormhole stabilization.

gr-qc↗

arXiv2025

High-dimensional Mean-Field Games by Particle-based Flow Matching

Mean-field games (MFGs) study the Nash equilibrium of systems with a continuum of interacting agents, which can be formulated as the fixed-point of optimal control problems. They provide a unified framework for a variety of applications, including optimal transport (OT) and generative models. Despite their broad applicability, solving high-dimensional MFGs remains a significant challenge due to fundamental computational and analytical obstacles. In this work, we propose a particle-based deep Flow Matching (FM) method to tackle high-dimensional MFG computation. In each iteration of our proximal fixed-point scheme, particles are updated using first-order information, and a flow neural network is trained to match the velocity of the sample trajectories in a simulation-free manner. Theoretically, in the optimal control setting, we prove that our scheme converges to a stationary point sublinearly, and upgrade to linear (exponential) convergence under additional convexity assumptions. Our proof uses FM to induce an Eulerian coordinate (density-based) from a Lagrangian one (particle-based), and this also leads to certain equivalence results between the two formulations for MFGs when the Eulerian solution is sufficiently regular. Our method demonstrates promising performance on non-potential MFGs and high-dimensional OT problems cast as MFGs through a relaxed terminal-cost formulation.

stat.ML↗

arXiv2025

Flow-based Conformal Prediction for Multi-dimensional Time Series

Time series prediction underpins a broad range of downstream tasks across many scientific domains. Recent advances and increasing adoption of black-box machine learning models for time series prediction highlight the critical need for uncertainty quantification. While conformal prediction has gained attention as a reliable uncertainty quantification method, conformal prediction for time series faces two key challenges: (1) \textbf{leveraging correlations in observations and non-conformity scores to overcome the exchangeability assumption}, and (2) \textbf{constructing prediction sets for multi-dimensional outcomes}. To address these challenges, we propose a novel conformal prediction method for time series using flow with classifier-free guidance. We provide coverage guarantees by establishing exact non-asymptotic marginal coverage and a finite-sample bound on conditional coverage for the proposed method. Evaluations on real-world time series datasets demonstrate that our method constructs significantly smaller prediction sets than existing conformal prediction methods, maintaining target coverage.

cs.LG↗

arXiv2024

Inverse Probability of Treatment Weighting with Deep Sequence Models Enables Accurate treatment effect Estimation from Electronic Health Records

Observational data have been actively used to estimate treatment effect, driven by the growing availability of electronic health records (EHRs). However, EHRs typically consist of longitudinal records, often introducing time-dependent confoundings that hinder the unbiased estimation of treatment effect. Inverse probability of treatment weighting (IPTW) is a widely used propensity score method since it provides unbiased treatment effect estimation and its derivation is straightforward. In this study, we aim to utilize IPTW to estimate treatment effect in the presence of time-dependent confounding using claims records. Previous studies have utilized propensity score methods with features derived from claims records through feature processing, which generally requires domain knowledge and additional resources to extract information to accurately estimate propensity scores. Deep sequence models, particularly recurrent neural networks and self-attention-based architectures, have demonstrated good performance in modeling EHRs for various downstream tasks. We propose that these deep sequence models can provide accurate IPTW estimation of treatment effect by directly estimating the propensity scores from claims records without the need for feature processing. We empirically demonstrate this by conducting comprehensive evaluations using synthetic and semi-synthetic datasets.

cs.LG↗

arXiv2024

Transformer Conformal Prediction for Time Series

We present a conformal prediction method for time series using the Transformer architecture to capture long-memory and long-range dependencies. Specifically, we use the Transformer decoder as a conditional quantile estimator to predict the quantiles of prediction residuals, which are used to estimate the prediction interval. We hypothesize that the Transformer decoder benefits the estimation of the prediction interval by learning temporal dependencies across past prediction residuals. Our comprehensive experiments using simulated and real data empirically demonstrate the superiority of the proposed method compared to the existing state-of-the-art conformal prediction methods.

cs.LG↗

arXiv2024

Neural network-based CUSUM for online change-point detection

Change-point detection, detecting an abrupt change in the data distribution from sequential data, is a fundamental problem in statistics and machine learning. CUSUM is a popular statistical method for online change-point detection due to its efficiency from recursive computation and constant memory requirement, and it enjoys statistical optimality. CUSUM requires knowing the precise pre- and post-change distribution. However, post-change distribution is usually unknown a priori since it represents anomaly and novelty. Classic CUSUM can perform poorly when there is a model mismatch with actual data. While likelihood ratio-based methods encounter challenges facing high dimensional data, neural networks have become an emerging tool for change-point detection with computational efficiency and scalability. In this paper, we introduce a neural network CUSUM (NN-CUSUM) for online change-point detection. We also present a general theoretical condition when the trained neural networks can perform change-point detection and what losses can achieve our goal. We further extend our analysis by combining it with the Neural Tangent Kernel theory to establish learning guarantees for the standard performance metrics, including the average run length (ARL) and expected detection delay (EDD). The strong performance of NN-CUSUM is demonstrated in detecting change-point in high-dimensional data using both synthetic and real-world data.

cs.LG↗

arXiv2023

Federated Learning in Temporal Heterogeneity

In this work, we explored federated learning in temporal heterogeneity across clients. We observed that global model obtained by \texttt{FedAvg} trained with fixed-length sequences shows faster convergence than varying-length sequences. We proposed methods to mitigate temporal heterogeneity for efficient federated learning based on the empirical observation.

cs.LG↗

arXiv2023

Deep Attention Q-Network for Personalized Treatment Recommendation

Tailoring treatment for individual patients is crucial yet challenging in order to achieve optimal healthcare outcomes. Recent advances in reinforcement learning offer promising personalized treatment recommendations; however, they rely solely on current patient observations (vital signs, demographics) as the patient's state, which may not accurately represent the true health status of the patient. This limitation hampers policy learning and evaluation, ultimately limiting treatment effectiveness. In this study, we propose the Deep Attention Q-Network for personalized treatment recommendations, utilizing the Transformer architecture within a deep reinforcement learning framework to efficiently incorporate all past patient observations. We evaluated the model on real-world sepsis and acute hypotension cohorts, demonstrating its superiority to state-of-the-art models. The source code for our model is available at https://github.com/stevenmsm/RL-ICU-DAQN.

cs.LG↗

arXiv2020

Higgs, top and electro-weak precision measurements at future $e^+ e^-$ colliders; a combined effective field theory analysis with renormalization mixing

This paper presents a combined analysis of the potential of a future electron-positron collider to constrain the Higgs, top and electro-weak (EW) sectors of the Standard Model Effective Field Theory (SMEFT). The leading contributions of operators involving top quarks arise mostly at one-loop suppressed order and can be captured by the renormalization group mixing with Higgs operators. We perform global fits with an extended basis of 29 parameters, including both Higgs and top operators, to the projections for the Higgs, top and electro-weak precision measurements at the International Linear Collider (ILC). The determination of the Higgs boson couplings in the 250 GeV stage of the ILC is initially severely degraded by the additional top-quark degrees of freedom, but can be nearly completely recovered by the inclusion of precise measurements of top-quark EW couplings at the LHC. The physical Higgs couplings are relatively robust, as the top mass is larger than the energy scale of EW processes. The effect of the top operators on the bounds on the Wilson coefficients is much more pronounced and may limit our ability to identify the source of deviations from the Standard Model. Robust global bounds on all Wilson coefficients are only obtained when the 500 GeV stage of the ILC is included.

hep-ph↗

arXiv2020

Study of the $hγZ$ coupling at the ILC

We study the $h γZ$ coupling, which is a loop induced coupling in the Standard Model (SM), to probe new physics. In a global fit based on the SM Effective Field Theory, measurement of the SM $h γZ$ coupling can provide a very useful constraint, in particular for the precise determination of $hZZ$ and $hWW$ couplings. At the International Linear Collider (ILC), there are two direct ways to study the $h γZ$ coupling: one is to measure the branching ratio of the $h \to γZ$ decay and the other to measure the cross section for the $e^+e^- \to h γ$ process. We have performed a full simulation study of the $e^+e^- \to h γ$ process at the 250 GeV ILC, assuming 2 ab$^{-1}$ data collected by the International Large Detector (ILD). The expected 1$σ$ bound on the effective $hγZ$ coupling ($ζ_{AZ}$) combining measurements of the cross section for $e^+e^- \to h γ$ followed by $h \to b \bar{b}$ and the $h \to γZ$ branching ratio is $-0.0015<ζ_{AZ}<0.0015$. The expected significance for the signal cross section in the fully hadronic $h \to WW^*$ channel is 0.09 $σ$ for beam polarizations of $P(e^-,e^+)=(-80\%,+30\%)$.

hep-ex↗

arXiv2019

Epistasis-based Basis Estimation Method for Simplifying the Problem Space of an Evolutionary Search in Binary Representation

An evolutionary search space can be smoothly transformed via a suitable change of basis; however, it can be difficult to determine an appropriate basis. In this paper, a method is proposed to select an optimum basis can be used to simplify an evolutionary search space in a binary encoding scheme. The basis search method is based on a genetic algorithm and the fitness evaluation is based on the epistasis, which is an indicator of the complexity of a genetic algorithm. Two tests were conducted to validate the proposed method when applied to two different evolutionary search problems. The first searched for an appropriate basis to apply, while the second searched for a solution to the test problem. The results obtained after the identified basis had been applied were compared to those with the original basis, and it was found that the proposed method provided superior results.

cs.NE↗

arXiv2019

Simplifying the vacuum bazooka

This paper provides a simplified explanation of the vacuum bazooka through diagrams and builds a theoretical model only using concepts found in introductory mechanics. Our theory suggests that the velocity of the projectile is proportional to the hyperbolic tangent of time, and experimental measurements support this claim. We also find that the vacuum bazooka could be used to demonstrate the concept of terminal velocity in a classroom setting.

physics.ed-ph↗

arXiv2019

Study of the $hγZ$ coupling using $e^+e^-\xrightarrow{}h γ$ at the ILC

We study the $e^+e^-\to h γ$ process at the International Linear Collider (ILC) to probe new physics in the $hγZ$ coupling. The study is performed at a center of mass energy of 250 GeV and is based on the full simulation of the International Large Detector (ILD). The expected signal significance is found to be 0.53 $σ$ for an integrated luminosity of 2000 fb$^{-1}$ in the case of the standard model. The corresponding 95 $\%$ confidence level upper limit for the signal cross section is 1.08 fb for left-handed beam polarization.

hep-ex↗

arXiv2018

The effect of liquid on the vibrational intensity of a wineglass at steady state resonance

As a liquid is inserted into a wineglass, the natural frequency of the wineglass decreases. This phenomenon, known as pitch lowering, is well explained in past papers. However, previous literature have not yet mentioned that pitch lowering also reduces the resonance intensity of a wineglass. Thus, this present paper aims to extend the body of research on this topic by describing the relationship between pitch lowering and its effect on resonation intensity. To do so, we identify the vibrating wineglass wall as a damped harmonic oscillator, derive a theoretical model, and find that the resonance intensity of the wineglass is proportional to the square of its natural frequency, under the assumption that damping stays constant. However, our experiments showed the coefficient of damping to increase with respect to the amount of liquid, which caused the data to deviate from its theoretical predictions. We conclude by discussing the accuracy and limitation of our proposed model.

physics.class-ph↗

arXiv2018

Measuring granular flow through a funnel with a force sensor

This paper presents an apparatus to measure the flow of granular matter through a funnel with a force sensor. In addition to being easily reproducible, our proposed method enhances measurement accuracy from the popular method that uses a stopwatch. We notice, however, that our proposed setup induces an unwelcome force that confounds force sensor readings. This issue is discussed, and its significance on measurement accuracy is evaluated through theoretical and experimental work.

cond-mat.soft↗

arXiv2017

Spin glass behavior in frustrated quantum spin system CuAl2O4 with a possible orbital liquid state

CuAl2O4 is a normal spinel oxide having quantum spin, S=1/2 for Cu2+. It is a rather unique feature that the Cu2+ ions of CuAl2O4 sit at a tetrahedral position, not like the usual octahedral position for many oxides. At low temperatures, it exhibits all the thermodynamic evidence of a quantum spin glass. For example, the polycrystalline CuAl2O4 shows a cusp centered at ~2 K in the low-field dc magnetization data and a clear frequency dependence in the ac magnetic susceptibility while it displays logarithmic relaxation behavior in a time dependence of the magnetization. At the same time, there is a peak at ~2.3 K in the heat capacity, which shifts towards higher temperature with magnetic fields. On the other hand, there is no evidence of new superlattice peaks in the high-resolution neutron powder diffraction data when cooled from 40 to 0.4 K. This implies that there is no long-ranged magnetic order down to 0.4 K, thus confirming a spin glass-like ground state for CuAl2O4. Interestingly, there is no sign of structural distortion either although Cu2+ is a Jahn-Teller active ion. Thus, we claim that an orbital liquid state is the most likely ground state in CuAl2O4. Of further interest, it also exhibits a large frustration parameter, f = Theta_CW/Tm ~67, one of the largest values reported for spinel oxides. Our observations suggest that CuAl2O4 should be a rare example of a frustrated quantum spin glass with a good candidate for an orbital liquid state.

cond-mat.str-el↗