arXiv subjects

Won Chang

Publications and source records attributed to Won Chang.

At least 19 recordsLinked to original sources

arXiv2026

Scalable, Likelihood-Free Calibration of Ice-Sheet Models with Deep Diffusion Emulators and Feature Matching

The Antarctic ice sheet is a major source of uncertainty in future sea-level projections, and physical simulators such as the PSU3D-ICE model are essential for studying its evolution. Calibrating them against observations is challenging: the simulator outputs and observed ice-thickness fields are high-dimensional, spatially dependent, and semi-continuous, with a large point mass at zero denoting ice-free regions. These features make conventional Gaussian-process emulation and likelihood-based calibration ill-suited and computationally infeasible at full resolution. We propose the sequential calibration method guided by a diffusion model and a Siamese network (SC-DS), a fully neural framework. For emulation, we develop a single conditional diffusion model that jointly generates the binary ice presence--absence pattern and the continuous thickness field, using global and local conditioning to represent how the input parameters shape the output. Because the emulator induces an intractable likelihood, our likelihood-free calibration replaces the hand-chosen distance and tolerance of approximate Bayesian computation with a probabilistic acceptance rule learned by an iteratively retrained Siamese network, together with a data--model discrepancy adjustment. Applied to the West Antarctic Ice Sheet, SC-DS matches the accuracy of state-of-the-art Gaussian-process calibration at a fraction of its computational cost and scales to the full-resolution domain, where existing methods become intractable.

stat.AP↗

arXiv2026

An adaptive split-combine Gaussian mixture filter for nonlinear and multimodal state estimation

Filtering combines model predictions with measurements to estimate the probability density function (PDF) of a system state over time. The PDF often becomes highly asymmetric and even multimodal in nonlinear systems with oscillatory or chaotic dynamics. Such non-Gaussian features violate the single-Gaussian assumption underlying Kalman-type filters. To address this problem, Gaussian mixture filtering has been proposed. However, accurately propagating mixture components and adaptively adjusting their number and weights over time remain open challenges. Here, we develop an adaptive split-combine Gaussian mixture filter (AMF) that estimates the time evolution of asymmetric and multimodal PDFs by adaptively splitting and combining Gaussian particles without auxiliary online numerical optimization. Notably, the proposed splitting method guarantees a reduction in variance along a target level-set-point direction of a Gaussian particle. This enables accurate and efficient propagation of particles. We show that AMF consistently outperforms various baseline filters across diverse benchmarks, including single and coupled slow-fast Van der Pol oscillators and the Lorenz attractor. We also propose a parallel implementation of AMF, allowing high-fidelity PDF estimation with practical computational cost.

math.NA↗

arXiv2026

Synthetic Data Generation With Incomplete Survey Data Under Informative Sampling

We propose a Bayesian framework for data synthesis and imputation in complex survey settings with informative sampling. To address variance underestimation in existing Bayesian approaches and to accommodate the missing data encountered in survey data, we introduce an adaptive weighting scheme for parameter estimation. We show that the proposed weighting yields consistent estimators with an asymptotically valid Godambe information matrix. The framework is flexible, accommodating a broad class of Bayesian models and facilitating practical implementation. Simulation studies demonstrate that the proposed method provides accurate uncertainty quantification for both model parameters and synthetic population inference.

stat.ME↗

arXiv2026

Rethinking Forward Processes for Score-Based Nonlinear Data Assimilation in High Dimensions

Data assimilation is the process of estimating the state of a dynamical system over time by combining model predictions with measurements. This task becomes challenging when the system is nonlinear and high-dimensional. To address this, score-based Bayesian filters have recently emerged. However, these methods still show unsatisfactory performance in certain cases, particularly under spatially sparse measurements. Such degradation stems from heuristic approximations of the likelihood score, whose errors can accumulate over time. This limitation arises because the methods simply adopt a classical forward process for generative modeling that transforms a data distribution toward a Gaussian distribution, which is independent of the measurement equation. Here, we propose a forward process tailored for filtering that transforms the system state toward the measurement space, enabling a theoretically sound formulation of the likelihood score. Based on this, we develop the Measurement-Aware Score-based Filter (MASF). We evaluate MASF on Kolmogorov flow, a high-dimensional fluid benchmark with up to $\mathcal{O}(10^5)$ dimensions, under diverse measurement operators, including nonlinear cases with a dimensional mismatch between the state and the measurements. MASF shows improved performance over existing score-based filters and ensemble-type Kalman filters. Notably, MASF achieves up to a $28.2\times$ wall-clock speedup compared with the baselines when using amortized pretraining. Our implementation is available at \texttt{https://github.com/tcnllab-oss/masf}.

stat.ML↗

arXiv2025

Stick-Breaking Mixture Normalizing Flows with Component-Wise Tail Adaptation for Variational Inference

Normalizing flows with a Gaussian base provide a computationally efficient way to approximate posterior distributions in Bayesian inference, but they often struggle to capture complex posteriors with multimodality and heavy tails. We propose a stick-breaking mixture base with component-wise tail adaptation (StiCTAF) for posterior approximation. The method first learns a flexible mixture base to mitigate the mode-seeking bias of reverse KL divergence through a weighted average of component-wise ELBOs. It then estimates local tail indices of unnormalized densities and finally refines each mixture component using a shared backbone combined with component-specific tail transforms calibrated by the estimated indices. This design enables accurate mode coverage and anisotropic tail modeling while retaining exact density evaluation and stable optimization. Experiments on synthetic posteriors demonstrate improved tail recovery and better coverage of multiple modes compared to benchmark models. We also present a real-data analysis illustrating the practical benefits of our approach for posterior inference.

stat.ML↗

arXiv2025

A Spatio-Temporal Dirichlet Process Mixture Model on Linear Networks for Crime Data

Analyzing crime events is crucial to understand crime dynamics and it is largely helpful for constructing prevention policies. Point processes specified on linear networks can provide a more accurate description of crime incidents by considering the geometry of the city. We propose a spatio-temporal Dirichlet process mixture model on a linear network to analyze crime events in Valencia, Spain. We propose a Bayesian hierarchical model with a Dirichlet process prior to automatically detect space-time clusters of the events and adopt a convolution kernel estimator to account for the network structure in the city. From the fitted model, we provide crime hotspot visualizations that can inform social interventions to prevent crime incidents. Furthermore, we study the relationships between the detected cluster centers and the city's amenities, which provides an intuitive explanation of criminal contagion.

stat.AP↗

arXiv2023

Statistical Power Analysis for Designing Bulk, Single-Cell, and Spatial Transcriptomics Experiments: Review, Tutorial, and Perspectives

Gene expression profiling technologies have been used in various applications such as cancer biology. The development of gene expression profiling has expanded the scope of target discovery in transcriptomic studies, and each technology produces data with distinct characteristics. In order to guarantee biologically meaningful findings using transcriptomic experiments, it is important to consider various experimental factors in a systematic way through statistical power analysis. In this paper, we review and discuss the power analysis for three types of gene expression profiling technologies from a practical standpoint, including bulk RNA-seq, single-cell RNA-seq, and high-throughput spatial transcriptomics. Specifically, we describe the existing power analysis tools for each research objective for each of the bulk RNA-seq and scRNA-seq experiments, along with recommendations. On the other hand, since there are no power analysis tools for high-throughput spatial transcriptomics at this point, we instead investigate the factors that can influence power analysis.

q-bio.GN↗

arXiv2022

Fast Computer Model Calibration using Annealed and Transformed Variational Inference

Computer models play a crucial role in numerous scientific and engineering domains. To ensure the accuracy of simulations, it is essential to properly calibrate the input parameters of these models through statistical inference. While Bayesian inference is the standard approach for this task, employing Markov Chain Monte Carlo methods often encounters computational hurdles due to the costly evaluation of likelihood functions and slow mixing rates. Although variational inference (VI) can be a fast alternative to traditional Bayesian approaches, VI has limited applicability due to boundary issues and local optima problems. To address these challenges, we propose flexible VI methods based on deep generative models that do not require parametric assumptions on the variational distribution. We embed a surjective transformation in our framework to avoid posterior truncation at the boundary. Additionally, we provide theoretical conditions that guarantee the success of the algorithm. Furthermore, our temperature annealing scheme can prevent being trapped in local optima through a series of intermediate posteriors. We apply our method to infectious disease models and a geophysical model, illustrating that the proposed method can provide fast and accurate inference compared to its competitors.

stat.AP↗

arXiv2022

A Bayesian Convolutional Neural Network-based Generalized Linear Model

Convolutional neural networks (CNNs) provide flexible function approximations for a wide variety of applications when the input variables are in the form of images or spatial data. Although CNNs often outperform traditional statistical models in prediction accuracy, statistical inference, such as estimating the effects of covariates and quantifying the prediction uncertainty, is not trivial due to the highly complicated model structure and overparameterization. To address this challenge, we propose a new Bayesian approach by embedding CNNs within the generalized linear models (GLMs) framework. We use extracted nodes from the last hidden layer of CNN with Monte Carlo (MC) dropout as informative covariates in GLM. This improves accuracy in prediction and regression coefficient inference, allowing for the interpretation of coefficients and uncertainty quantification. By fitting ensemble GLMs across multiple realizations from MC dropout, we can account for uncertainties in extracting the features. We apply our methods to biological and epidemiological problems, which have both high-dimensional correlated inputs and vector covariates. Specifically, we consider malaria incidence data, brain tumor image data, and fMRI data. By extracting information from correlated inputs, the proposed method can provide an interpretable Bayesian analysis. The algorithm can be broadly applicable to image regressions or correlated data analysis by enabling accurate Bayesian inference quickly.

stat.ME↗

arXiv2022

A Spatio-Temporal Dirichlet Process Mixture Model for Coronavirus Disease-19

Understanding the spatio-temporal patterns of the coronavirus disease 2019 (COVID-19) is essential to construct public health interventions. Spatially referenced data can provide richer opportunities to understand the mechanism of the disease spread compared to the more often encountered aggregated count data. We propose a spatio-temporal Dirichlet process mixture model to analyze confirmed cases of COVID-19 in an urban environment. Our method can detect unobserved cluster centers of the epidemics, and estimate the space-time range of the clusters that are useful to construct a warning system. Furthermore, our model can measure the impact of different types of landmarks in the city, which provides an intuitive explanation of disease spreading sources from different time points. To efficiently capture the temporal dynamics of the disease patterns, we employ a sequential approach that uses the posterior distribution of the parameters for the previous time step as the prior information for the current time step. This approach enables us to incorporate time dependence into our model in a computationally efficient manner without complicating the model structure. We also develop a model assessment by comparing the data with theoretical densities, and outline the goodness-of-fit of our fitted model.

stat.ME↗

arXiv2022

graph-GPA 2.0: A Graphical Model for Multi-disease Analysis of GWAS Results with Integration of Functional Annotation Data

Genome-wide association studies (GWAS) have successfully identified a large number of genetic variants associated with traits and diseases. However, it still remains challenging to fully understand functional mechanisms underlying many associated variants. This is especially the case when we are interested in variants shared across multiple phenotypes. To address this challenge, we propose graph-GPA 2.0 (GGPA 2.0), a novel statistical framework to integrate GWAS datasets for multiple phenotypes and incorporate functional annotations within a unified framework. We conducted simulation studies to evaluate GGPA 2.0. The results indicate that incorporating functional annotation data using GGPA 2.0 does not only improve detection of disease-associated variants, but also allows to identify more accurate relationships among diseases. We analyzed five autoimmune diseases and five psychiatric disorders with the functional annotations derived from GenoSkyline and GenoSkyline-Plus and the prior disease graph generated by biomedical literature mining. For autoimmune diseases, GGPA 2.0 identified enrichment for blood, especially B cells and regulatory T cells across multiple diseases. Psychiatric disorders were enriched for brain, especially prefrontal cortex and inferior temporal lobe for bipolar disorder (BIP) and schizophrenia (SCZ), respectively. Finally, GGPA 2.0 successfully identified the pleiotropy between BIP and SCZ. These results demonstrate that GGPA 2.0 can be a powerful tool to identify associated variants associated with each phenotype or those shared across multiple phenotypes, while also promoting understanding of functional mechanisms underlying the associated variants.

q-bio.GN↗

arXiv2021

Bayesian Model Calibration and Sensitivity Analysis for Oscillating Biological Experiments

Understanding the oscillating behaviors that govern organisms' internal biological processes requires interdisciplinary efforts combining both biological and computer experiments, as the latter can complement the former by simulating perturbed conditions with higher resolution. Harmonizing the two types of experiment, however, poses significant statistical challenges due to identifiability issues, numerical instability, and ill behavior in high dimension. This article devises a new Bayesian calibration framework for oscillating biochemical models. The proposed Bayesian model is estimated relying on an advanced Markov chain Monte Carlo (MCMC) technique which can efficiently infer the parameter values that match the simulated and observed oscillatory processes. Also proposed is an approach to sensitivity analysis based on the intervention posterior. This approach measures the influence of individual parameters on the target process by using the obtained MCMC samples as a computational tool. The proposed framework is illustrated with circadian oscillations observed in a filamentous fungus, Neurospora crassa.

stat.AP↗

arXiv2021

Fast and accurate learned multiresolution dynamical downscaling for precipitation

This study develops a neural network-based approach for emulating high-resolution modeled precipitation data with comparable statistical properties but at greatly reduced computational cost. The key idea is to use combination of low- and high- resolution simulations to train a neural network to map from the former to the latter. Specifically, we define two types of CNNs, one that stacks variables directly and one that encodes each variable before stacking, and we train each CNN type both with a conventional loss function, such as mean square error (MSE), and with a conditional generative adversarial network (CGAN), for a total of four CNN variants. We compare the four new CNN-derived high-resolution precipitation results with precipitation generated from original high resolution simulations, a bilinear interpolater and the state-of-the-art CNN-based super-resolution (SR) technique. Results show that the SR technique produces results similar to those of the bilinear interpolator with smoother spatial and temporal distributions and smaller data variabilities and extremes than the original high resolution simulations. While the new CNNs trained by MSE generate better results over some regions than the interpolator and SR technique do, their predictions are still not as close as the original high resolution simulations. The CNNs trained by CGAN generate more realistic and physically reasonable results, better capturing not only data variability in time and space but also extremes such as intense and long-lasting storms. The new proposed CNN-based downscaling approach can downscale precipitation from 50~km to 12~km in 14~min for 30~years once the network is trained (training takes 4~hours using 1~GPU), while the conventional dynamical downscaling would take 1~month using 600 CPU cores to generate simulations at the resolution of 12~km over contiguous United States.

cs.LG↗

arXiv2020

Computer Model Calibration with Time Series Data using Deep Learning and Quantile Regression

Computer models play a key role in many scientific and engineering problems. One major source of uncertainty in computer model experiment is input parameter uncertainty. Computer model calibration is a formal statistical procedure to infer input parameters by combining information from model runs and observational data. The existing standard calibration framework suffers from inferential issues when the model output and observational data are high-dimensional dependent data such as large time series due to the difficulty in building an emulator and the non-identifiability between effects from input parameters and data-model discrepancy. To overcome these challenges we propose a new calibration framework based on a deep neural network (DNN) with long-short term memory layers that directly emulates the inverse relationship between the model output and input parameters. Adopting the 'learning with noise' idea we train our DNN model to filter out the effects from data model discrepancy on input parameter inference. We also formulate a new way to construct interval predictions for DNN using quantile regression to quantify the uncertainty in input parameter estimates. Through a simulation study and real data application with WRF-hydro model we show that our approach can yield accurate point estimates and well calibrated interval estimates for input parameters.

stat.ML↗

arXiv2019

Ice Model Calibration Using Semi-continuous Spatial Data

Rapid changes in Earth's cryosphere caused by human activity can lead to significant environmental impacts. Computer models provide a useful tool for understanding the behavior and projecting the future of Arctic and Antarctic ice sheets. However, these models are typically subject to large parametric uncertainties due to poorly constrained model input parameters that govern the behavior of simulated ice sheets. Computer model calibration provides a formal statistical framework to infer parameters using observational data, and to quantify the uncertainty in projections due to the uncertainty in these parameters. Calibration of ice sheet models is often challenging because the relevant model output and observational data take the form of semi-continuous spatial data, with a point mass at zero and a right-skewed continuous distribution for positive values. Current calibration approaches cannot handle such data. Here we introduce a hierarchical latent variable model that handles binary spatial patterns and positive continuous spatial patterns as separate components. To overcome challenges due to high-dimensionality we use likelihood-based generalized principal component analysis to impose low-dimensional structures on the latent variables for spatial dependence. We apply our methodology to calibrate a physical model for the Antarctic ice sheet and demonstrate that we can overcome the aforementioned modeling and computational challenges. As a result of our calibration, we obtain improved future ice-volume change projections.

stat.ME↗

arXiv2019

Computer model calibration based on image warping metrics: an application for sea ice deformation

Arctic sea ice plays an important role in the global climate. Sea ice models governed by physical equations have been used to simulate the state of the ice including characteristics such as ice thickness, concentration, and motion. More recent models also attempt to capture features such as fractures or leads in the ice. These simulated features can be partially misaligned or misshapen when compared to observational data, whether due to numerical approximation or incomplete physics. In order to make realistic forecasts and improve understanding of the underlying processes, it is necessary to calibrate the numerical model to field data. Traditional calibration methods based on generalized least-square metrics are flawed for linear features such as sea ice cracks. We develop a statistical emulation and calibration framework that accounts for feature misalignment and misshapenness, which involves optimally aligning model output with observed features using cutting edge image registration techniques. This work can also have application to other physical models which produce coherent structures.

stat.AP↗

arXiv2018

A Regularized Spatial Market Segmentation Method with Dirichlet Process Gaussian Mixture Prior

Spatially referenced data are increasingly available thanks to the development of modern GPS technology. They also provide rich opportunities for spatial analytics in the field of marketing science. Our main interest is to propose a new efficient statistical framework to conduct spatial segmentation analysis for restaurants located in a metropolitan area in the U.S. The spatial segmentation problem poses important statistical challenges: selecting the optimal number of underlying structures of market segments, capturing complex and flexible spatial structures, and resolving any possible small n and large p issue which can be typical in latent class analysis. Existing approaches try to tackle these issues in heuristic ways or seem silent on them. To overcome these challenges, we propose a new statistical framework based on regularized Bayesian spatial mixture regressions with Dirichlet process integrating ridge or lasso regularization. Our simulation study demonstrates that the proposed models successfully recover the underlying spatial clustering structures and outperforms two existing benchmark models. In the empirical analysis using online customer satisfaction data from the Yelp, our models provides interesting insights on segment-level key drivers of customer satisfaction and interpretable relationships between regional demographics and restaurants' characteristics.

stat.AP↗

arXiv2017

Diagnosing added value of convection-permitting regional models using precipitation event identification and tracking

Dynamical downscaling with high-resolution regional climate models may offer the possibility of realistically reproducing precipitation and weather events in climate simulations. As resolutions fall to order kilometers, the use of explicit rather than parametrized convection may offer even greater fidelity. However, these increased model resolutions both allow and require increasingly complex diagnostics for evaluating model fidelity. In this study we use a suite of dynamically downscaled simulations of the summertime U.S. (WRF driven by NCEP) with systematic variations in parameters and treatment of convection as a test case for evaluation of model precipitation. In particular, we use a novel rainstorm identification and tracking algorithm that allocates essentially all rainfall to individual precipitation events (Chang et al. 2016). This approach allows multiple insights, including that, at least in these runs, model wet bias is driven by excessive areal extent of precipitating events. Biases are time-dependent, producing excessive diurnal cycle amplitude. We show that this effect is produced not by new production of events but by excessive enlargement of long-lived precipitation events during daytime, and that in the domain average, precipitation biases appear best represented as additive offsets. Of all model configurations evaluated, convection-permitting simulations most consistently reduced biases in precipitation event characteristics.

stat.AP↗