A5001163175- Statistical Methods and Inference
- Bayesian Methods and Mixture Models
- Soil Geostatistics and Mapping
- Tensor decomposition and applications
- Gaussian Processes and Bayesian Inference
- Advanced Neuroimaging Techniques and Applications
- Statistical Methods and Bayesian Inference
- Sparse and Compressive Sensing Techniques
- Spatial and Panel Data Analysis
- Functional Brain Connectivity Studies
- Face and Expression Recognition
- Remote Sensing and LiDAR Applications
- Forest ecology and management
- Economic and Environmental Valuation
- Data Management and Algorithms
- Data Analysis with R
- Remote Sensing in Agriculture
- Neurobiology of Language and Bilingualism
- Computational Physics and Python Applications
- Vaccine Coverage and Hesitancy
- Advanced Statistical Methods and Models
- Brain Tumor Detection and Classification
- Land Use and Ecosystem Services
- Data-Driven Disease Surveillance
- Social Capital and Networks
Texas A&M University
2023-2024
University of California, Santa Cruz
2016-2021
Brigham Young University
2018
Duke University
2012-2014
St. John's National Academy of Health Sciences
2014
University of Minnesota
2011
The Gaussian process is an indispensable tool for spatial data analysts. onset of the "big data" era, however, has lead to traditional being computationally infeasible modern data. As such, various alternatives full that are more amenable handling big have been proposed. These methods often exploit low-rank structures and/or multi-core and multi-threaded computing environments facilitate computation. This study provides, first, introductory overview several analyzing large Second, this...
We propose a Bayesian approach to regression with scalar response on vector and tensor covariates. Vectorization of the prior analysis fails exploit structure, often leading p...
Spatial process models for analyzing geostatistical data entail computations that become prohibitive as the number of spatial locations becomes large. There is a burgeoning literature on approaches large datasets. In this article, we propose divide-and-conquer strategy within Bayesian paradigm. We partition into subsets, analyze each subset using model and then obtain approximate posterior inference entire dataset by combining individual distributions from subset. Importantly, often desired...
As an alternative to variable selection or shrinkage in high-dimensional regression, we propose randomly compress the predictors prior analysis. This dramatically reduces storage and computational bottlenecks, performing well when can be projected a low-dimensional linear subspace with minimal loss of information about response. opposed existing Bayesian dimensionality reduction approaches, exact posterior distribution conditional on compressed data is available analytically, speeding up...
Large point referenced datasets occur frequently in the environmental and natural sciences. Use of Bayesian hierarchical spatial models for analyzing these is undermined by onerous computational burdens associated with parameter estimation. Low-rank process attempt to resolve this problem projecting effects a lower-dimensional subspace. This subspace determined judicious choice "knots" or locations that are fixed priori. One such representation yields class predictive (e.g., Banerjee et al.,...
Developing surrogates for computer models has become increasingly important addressing complex problems in science and engineering. This article introduces an artificial intelligent (AI) surrogate, referred to as the DeepSurrogate, analyzing functional outputs with vector-valued inputs. The relationship between output input is modeled infinite sequence of unknown functions, each representing at a specific location within domain. These spatially indexed functions are expressed through...
Abstract Background Human papillomavirus ( HPV ) is the most common sexually transmitted infection in world. It can lead to anogenital, cervical, and head neck cancer, with higher risk of malignant disease patients human immunodeficiency virus HIV )/acquired syndrome AIDS patients. In India, 73,000 130,000 women diagnosed cervical cancer die annually. Gardasil ® , a vaccine available against types 6, 11, 16, 18, approved for use India but not men. A backlash post‐licensure trials has created...
We propose a three-step divide-and-conquer strategy within the Bayesian paradigm that delivers massive scalability for any spatial process model. partition data into large number of subsets, apply readily available model on every subset, in parallel, and optimally combine posterior distributions estimated across all subsets pseudo-posterior distribution conditions entire data. The combined pseudo replaces full predicting responses at arbitrary locations inference parameters surface. Based...
This article proposes a novel Bayesian implementation of regression with multi-dimensional array (tensor) response on scalar covariates. The recent emergence complex datasets in various disciplines presents pressing need to devise models tensor valued response. considers one such application detecting neuronal activation fMRI experiments presence brain images and predictors. overarching goal this is identify spatial regions (voxels) activated by an external stimulus. In related applications,...
Nonparametric regression for large numbers of features (p) is an increasingly important problem. If the sample size n massive, a common strategy to partition feature space, and then separately apply simple models each set. This not ideal when modest relative p, we propose alternative approach relying on random compression vector combined with Gaussian process regression. The proposed particularly motivated by setting in which response conditionally independent given projection low...
Gaussian process (GP) regression is computationally expensive in spatial applications involving massive data. Various methods address this limitation, including a small number of Bayesian based on distributed computations (or the divide-and-conquer strategy). Focusing latter literature, we achieve three main goals. First, develop an extensible framework for GP that embeds many popular methods. The proposed has steps partition entire data into subsets, apply readily available model parallel...
This article proposes a framework based on shared, time varying stochastic latent factor models for modeling relational data in which network and node-attributes co-evolve over time. Our proposed is flexible enough to handle both categorical continuous attributes, allows us estimate the dimension of social space, automatically yields Bayesian hypothesis tests association between structure nodal attributes. Additionally, model easy compute readily inference prediction missing link nodes. We...
This article focuses on model-based clustering of subjects based the shared relationships subject-specific networks and covariates in scenarios when there are differences relationship between for different groups subjects. It is also interest to identify network nodes significantly associated with each covariate cluster To address these methodological questions, we propose a novel nonparametric Bayesian mixture modeling framework an undirected response scalar predictors. The symmetric matrix...
Motivated by brain connectome datasets acquired using diffusion weighted magnetic resonance imaging (DWI), this article proposes a novel generalized Bayesian linear modeling framework with symmetric tensor response and scalar predictors. The coefficients corresponding to the predictors are embedded two features: low-rankness group sparsity within low-rank structure. Besides offering computational efficiency parsimony, these features enable identification of important "tensor nodes" cells"...
We propose a conditional density filtering (C-DF) algorithm for efficient online Bayesian inference. C-DF adapts MCMC sampling to the setting, from approximations posterior distributions obtained by propagating surrogate sufficient statistics (a function of data and parameter estimates) as new arrive. These quantities eliminate need store or process entire dataset simultaneously offer number desirable features. Often, these include reduction in memory requirements runtime improved mixing,...
Brain activation and connectivity analyses in task-based functional magnetic resonance imaging (fMRI) experiments with multiple subjects are currently at the forefront of data-driven neuroscience. In such experiments, interest often lies understanding brain voxels due to external stimuli strong association or between measurements on a set pre-specified group voxels, also known as regions (ROI). This article proposes joint Bayesian additive mixed modeling framework that simultaneously...