Recent advances in Bayesian analysis for complex datasets

1775 

Topic-Contributed Paper Session 

Sayantan Banerjee (1), Subhashis Ghoshal (2), Weining Shen (3), Jyotishka Datta (4), Yuan Wang (5), Maoran Xu (6), Tianyu Pan (7)

(1) Indian Institute of Management Indore, N/A, (2) North Carolina State University, N/A, (3) University of California, Irvine, N/A, (4) Virginia Tech, N/A, (5) Washington State University, N/A, (6) Duke University, N/A, (7) Stanford University, N/A

Subhashis Ghoshal  
North Carolina State University

Sayantan Banerjee  
Indian Institute of Management Indore

Weining Shen  
University of California, Irvine

Jyotishka Datta  
Virginia Tech

Yuan Wang  
Washington State University

Maoran Xu  
Duke University

Tianyu Pan  
Stanford University

Description of the session and focus:
The proposed session aims to offer an in-depth exploration of Bayesian methods as a robust and flexible paradigm for addressing the intricate challenges inherent in complex data analysis. As data sources continue to evolve in their sophistication and complexity, Bayesian techniques provide a principled approach to modelling uncertainty, extracting hidden structures, and drawing accurate inferences. This session will delve into advanced Bayesian methodologies and their practical applications across various domains.

Content:
The session will explore the modern advances in Bayesian methods for complex data, for example, Bayesian methods for bi-clustering, non-linear independent component estimation, regression models for misreported responses, nonparametric factor models, selecting relevant external data, apart from exploring computational methods like variational inference, MCMC and Hamiltonian Monte Carlo.

Tentative titles:
Weining Shen – Bayesian bi-clustering and its application in education data analysis
Jyotishka Datta -- Deep Bayesian Non-linear Independent Component Estimation
Yuan Wang – A Bayesian Regression Model with Misreported Response
Maoran Xu – Identifiable and interpretable nonparametric factor analysis
Tianyu Pan -- A Bayesian Approach for Selecting Relevant External Data (BASE)

Timeliness:
The session on Bayesian methods for complex data is especially timely due to several critical factors:
(i) High-dimensional data: As datasets continue to grow in dimensionality, Bayesian techniques provide a coherent framework for dimensionality reduction, feature selection, and effective modeling.
(ii) Computational advances: Recent advancements in computational tools and hardware have made Bayesian methods more accessible for handling large-scale and complex datasets.
(iii) Interdisciplinary applications: Bayesian methods are increasingly adopted in diverse fields, such as bioinformatics, finance, and natural language processing, showcasing their versatility and effectiveness.
Appeal:
This session is designed for researchers, data scientists, statisticians, and practitioners with a solid foundation in Bayesian statistics and a desire to deepen their understanding of advanced Bayesian methodologies. Attendees will gain practical insights into implementing and customizing Bayesian models for complex data scenarios, enabling them to extract valuable insights and derive meaningful conclusions from challenging datasets.

No Additional Sponsor ³

No Additional Sponsor ²

International Indian Statistical Association ¹