Article
Engineering, Multidisciplinary
Nicholas Oune, Ramin Bostanabad
Summary: LMGPs are a novel type of Gaussian processes that can handle mixed data by encoding qualitative inputs in a continuous latent space. By systematically learning optimal mapping and corresponding manifold through maximum likelihood estimation, LMGPs outperform existing methods in terms of accuracy and versatility.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2021)
Article
Engineering, Electrical & Electronic
Zhaojing Wang, Ying Zheng, David Shan-Hill Wong, Yang Wang, Weidong Yang
Summary: This article introduces a generalized monitoring scheme for industrial processes, using local average similarity and distance average similarity to divide operational stages and identify repeating stages. The method also employs multiple cointegration analysis and detrended fluctuation regression models to handle nonstationary variables and map all stages into a stationary space, leading to effective process monitoring.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2022)
Article
Statistics & Probability
Omar Dahdouh, Majid Jafari Khaledi
Summary: This paper develops a Bayesian nonparametric model for skewed spatial data with nonstationary dependence structure. A transformed Gaussian model is proposed for the atoms of the kernel stick-breaking process by transforming the margins of a Gaussian process to flexible marginal distributions. This study proves that the correlation structure of the underlying spatial process is nonstationary. Results from both simulated and real datasets demonstrate that the proposed model possesses better spatial prediction performance and offers computational advantages compared to the Bayesian nonparametric model with the Gaussian base measure.
COMMUNICATIONS IN STATISTICS-SIMULATION AND COMPUTATION
(2022)
Article
Health Care Sciences & Services
Yusuke Saigusa, Shinto Eguchi, Osamu Komori
Summary: The generalized linear mixed model (GLMM) is a common method for analyzing longitudinal and clustered data in biological sciences. However, issues of model complexity and misspecification can arise. This paper extends the standard GLMM to a nonlinear mixed-effects model based on quasi-linear modeling, providing an estimation algorithm and a conditional AIC for the proposed model. Performance under model misspecification is evaluated in simulation studies, and the proposed model is shown to capture heterogeneity in respiratory illness data.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2022)
Article
Statistics & Probability
Aritra Halder, Sudipto Banerjee, Dipak K. Dey
Summary: Spatial process models are commonly used for modeling point-referenced variables in different scientific domains. Analyzing random surfaces helps explore the latent dependence within the studied response. In this article, we present a Bayesian approach to model rapid changes on the response surface and assess directional curvature along a given trajectory. We demonstrate the effectiveness of our methodology using simulated experiments, as well as real-world applications involving Boston Housing data, Meuse river data, and temperature data from the Northeastern United States.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Computer Science, Interdisciplinary Applications
Italo Gomes Goncalves, Felipe Guadagnin, Diogo Peixoto Cordova
Summary: This paper introduces a variational Gaussian process (VGP) model specialized in spatial data by leveraging recent advances in the machine learning field. The model is highly modular and customizable, allowing for different assumptions about the data. The focus of this work is on multivariate robust regression using an adaptation of the e-insensitive loss function. The VGP model enables end-to-end modeling with normal score transformation, spatial pattern detection, and prediction. The paper also presents a methodology for handling large datasets and provides an open-source implementation.
COMPUTERS & GEOSCIENCES
(2022)
Article
Physics, Multidisciplinary
Vasiliki D. Agou, Andrew Pavlides, Dionissios T. Hristopulos
Summary: The study introduces a data-driven model for precipitation amount using warped Gaussian processes to predict spatiotemporal patterns of precipitation. The advantages of non-parametric warping for interpolation of incomplete data are established through cross-validation analysis. The model offers enhanced flexibility and improved predictive accuracy for non-Gaussian data.
Article
Biochemical Research Methods
Nuha BinTayyash, Sokratia Georgaka, S. T. John, Sumon Ahmed, Alexis Boukouvalas, James Hensman, Magnus Rattray
Summary: The GPcounts package implements GP regression methods for modeling counts data using a negative binomial likelihood function with computational efficiency achieved through variational Bayesian inference. The method shows better performance in identifying changes in over-dispersed counts data compared to methods based on Gaussian or Poisson likelihoods, as validated on simulated data.
Article
Automation & Control Systems
Jackson Loper, David Blei, John P. Cunningham, Liam Paninski
Summary: The study demonstrates that in one-dimensional space, using the Latent Exponentially Generated (LEG) family of state-space models can effectively approximate any one-dimensional Gaussian Processes, while offering more advantages. Inference and learning can be performed using parallelized algorithms, tested on real and synthetic data, and scaled to datasets with billions of samples.
JOURNAL OF MACHINE LEARNING RESEARCH
(2021)
Article
Water Resources
Yue Zhao, Jian Luo
Summary: This study introduces an iterative algorithm for large-scale spatial fields inverse problems that corrects biases in spatial covariance by updating dominant principal components. The method significantly improves inversion results for large-scale inverse problems with biased structural parameters for spatial covariance, matching distribution patterns of spatially correlated parameter field with dominant principal components.
ADVANCES IN WATER RESOURCES
(2021)
Article
Chemistry, Analytical
Yuchen Zou, Weiwei Tang, Bin Li
Summary: Spatial segmentation is important in mass spectrometry imaging (MSI) data analysis, as it helps to identify homogeneous/heterogeneous subgroups and provides vital characteristics for biological analysis. This study proposes a segmentation pipeline that utilizes pattern compression by principal component analysis (PCA) for easy and effective segmentation.
Article
Physics, Multidisciplinary
Frank Aurzada, Martin Kilian, Ercan Sonmez
Summary: In this paper, we investigate the asymptotic behavior of the persistence probability of the sum of two self-similar centered Gaussian processes with different self-similarity indices. Under the assumption of non-negative correlations and some additional minor conditions, we show that the asymptotic behavior of the persistence probability of the sum is equivalent to that of the process with the greater self-similarity index.
JOURNAL OF PHYSICS A-MATHEMATICAL AND THEORETICAL
(2022)
Article
Mathematics, Applied
Almudena P. Marquez, Francisco Javier Garcia-Pacheco, Miriam Mengibar-Rodriguez, Alberto Sanchez-Alzola
Summary: This manuscript introduces the concepts of bounded linear operators and supporting vectors, and explores the relationship between the principal components of a matrix and its supporting vectors.
Article
Public, Environmental & Occupational Health
Biruk Shalmeno Tusa, Adisu Birhanu Weldesenbet, Telahun Kasa Tefera, Sewnet Adem Kebede
Summary: The study analyzed the spatial distribution of male circumcision in Ethiopia using data from the 2016 Ethiopian Demographic and Health Survey. The distribution was found to be nonrandom, with primary clusters identified in regions such as Oromia and Tigray. Factors associated with male circumcision included age, age at circumcision, ethnicity, religion, and place of residence.
Article
Mathematics
Giacomo Ascione, Enrica Pirozzi
Summary: This paper focuses on constructing deterministic and stochastic extensions of the Gompertz curve using generalized fractional derivatives induced by complete Bernstein functions. It introduces a class of linear stochastic equations involving a generalized fractional integral and studies the properties of its solutions, proving the existence and uniqueness of Gaussian solutions via a fixed point argument and showing that expected value of the solution solves a generalized fractional linear equation under suitable conditions.
Article
Meteorology & Atmospheric Sciences
Won Chang, Jiali Wang, Julian Marohnic, V. Rao Kotamarthi, Elisabeth J. Moyer
Article
Biology
Yawen Guan, Christian Sampson, J. Derek Tucker, Won Chang, Anirban Mondal, Murali Haran, Deborah Sulsky
JOURNAL OF AGRICULTURAL BIOLOGICAL AND ENVIRONMENTAL STATISTICS
(2019)
Article
Geosciences, Multidisciplinary
Won Chang, Sunghoon Kim, Heewon Chae
SPATIAL STATISTICS
(2020)
Article
Statistics & Probability
Matthew Plumlee, Taylor G. Asher, Won Chang, Matthew Bilskie
Summary: This article demonstrates the good forecasting performance of probabilistic hurricane storm surge forecasting using a small number of carefully chosen simulations by utilizing modern statistical tools and an optimal design criterion. The study also addresses the missing data and output handling issues in surge modeling, showing evidence of efficacy in comparison to existing methods through a case study on Hurricane Michael (2018).
ANNALS OF APPLIED STATISTICS
(2021)
Article
Business
Sunghoon Kim, Wayne S. DeSarbo, Won Chang
Summary: The study introduces a new spatial modeling approach to calibrate the potential impact of spatial dependency and heterogeneity on customer service and satisfaction measurements. By utilizing a hierarchical Bayes framework with geographical boundary effects, the proposed method shows improved performance compared to existing procedures.
INTERNATIONAL JOURNAL OF RESEARCH IN MARKETING
(2021)
Article
Computer Science, Interdisciplinary Applications
Jacob Tracy, Won Chang, Sarah St George Freeman, Casey Brown, Adriana Palma Nava, Patrick Ray
Summary: This paper introduces a data-driven dynamic emulation method that successfully simulates high-dimensional model outputs, overcoming challenges from traditional multivariate statistics. The method is demonstrated on a regional groundwater model in metropolitan Mexico City and successfully emulates the dynamic of land subsidence and aquifer level fluctuation, providing new methodological advances for computationally expensive planning and optimization applications.
ENVIRONMENTAL MODELLING & SOFTWARE
(2022)
Article
Mathematics, Interdisciplinary Applications
Saumya Bhatnagar, Won Chang, Seonjin Kim, Jiali Wang
Summary: Computer model calibration plays a key role in scientific and engineering problems. However, the existing standard calibration framework faces inferential issues when dealing with high-dimensional dependent data. To overcome this problem, we propose a new calibration framework based on deep neural networks, which can accurately estimate input parameters and quantify the uncertainty of the estimates.
SIAM-ASA JOURNAL ON UNCERTAINTY QUANTIFICATION
(2022)
Article
Geosciences, Multidisciplinary
Jaewoo Park, Won Chang, Boseung Choi
Summary: This article analyzes COVID-19 contact tracing data from Seoul to understand the spatial patterns of patient visits and detect local cluster centers. The study develops a novel interaction Neyman-Scott process to account for the complex behavior of cluster centers. The results show that the method can effectively describe the spatial patterns of patient visits and provide visualizations for public health interventions.
SPATIAL STATISTICS
(2022)
Article
Biology
Carter Allen, Yuzhou Chang, Brian Neelon, Won Chang, Hang J. Kim, Zihai Li, Qin Ma, Dongjun Chung
Summary: High throughput spatial transcriptomics (HST) is a rapidly emerging experimental technology that allows for analyzing gene expression in tissue samples at or near single-cell resolution. We developed SPRUCE, a Bayesian spatial multivariate finite mixture model, to identify distinct cellular sub-populations in HST data.
Article
Statistics & Probability
Won Chang, Bledar A. Konomi, Georgios Karagiannis, Yawen Guan, Murali Haran
Summary: Calibrating ice sheet models is challenging due to the nature of the data and the uncertainties in model parameters. However, a hierarchical latent variable model can overcome these challenges and provide improved projections for future ice-volume change.
ANNALS OF APPLIED STATISTICS
(2022)
Article
Environmental Sciences
Christopher K. Wikle, Abhirup Datta, Bhava Vyasa Hari, Edward L. Boone, Indranil Sahoo, Indulekha Kavila, Stefano Castruccio, Susan J. Simmons, Wesley S. Burr, Won Chang
Summary: Explainable AI is a sub-discipline of computer science and machine learning that aims to address the lack of uncertainty quantification and inability to do inference in machine learning and deep neural models. This article focuses on explaining which inputs are important in models for predicting environmental data, and describes three general methods for explainability.
Review
Biochemistry & Molecular Biology
Hyeongseon Jeon, Juan Xie, Yeseul Jeon, Kyeong Joo Jung, Arkobrato Gupta, Won Chang, Dongjun Chung
Summary: In this paper, power analysis for three types of gene expression profiling technologies, bulk RNA-seq, single-cell RNA-seq, and high-throughput spatial transcriptomics, is reviewed and discussed from a practical standpoint. Existing power analysis tools and recommendations are described for bulk RNA-seq and scRNA-seq experiments. Factors influencing power analysis are investigated for high-throughput spatial transcriptomics, as there are currently no power analysis tools available.
Article
Environmental Sciences
Jaehong Jeong, Won Chang
Summary: As the risk of climate change becomes more apparent, countries worldwide are actively searching for alternative energy sources. Wind energy presents great potential for future energy portfolios without negative environmental impacts. Understanding the spatio-temporal patterns of wind is crucial in developing energy plans at national and global levels.
Article
Clinical Neurology
Minki P. Lee, Kien Hoang, Sungkyu Park, Yun Min Song, Eun Yeon Joo, Won Chang, Jee Hyun Kim, Jae Kyoung Kim
Summary: Sleep is crucial for health and well-being, but collecting and analyzing accurate longitudinal sleep data can be challenging. Researchers propose a neural network model called SOMNI to impute missing sleep-activity data, allowing clinicians to monitor sleep-wake cycles of individuals with irregular sleep patterns.
Article
Geosciences, Multidisciplinary
Jiali Wang, Zhengchun Liu, Ian Foster, Won Chang, Rajkumar Kettimuthu, V. Rao Kotamarthi
Summary: This study develops a neural-network-based approach for emulating high-resolution precipitation data with greatly reduced computational cost, which trains on a combination of low- and high-resolution simulations to produce realistic results. Results show that CNNs trained by CGAN generated more physically reasonable results, better capturing data variability and extremes than conventional methods.
GEOSCIENTIFIC MODEL DEVELOPMENT
(2021)