Bayesian Image Regression with Soft-thresholded Conditional Autoregressive Prior

Authors: Yuliang Xu, Jian Kang

ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Simulation studies demonstrate that the ST-CAR prior outperforms existing methods in identifying active brain regions with complex correlation patterns, while our VI algorithms offer superior computational performance. We further validate our approach by applying the ST-CAR to working memory f MRI data from the Adolescent Brain Cognitive Development (ABCD) study, highlighting its effectiveness in practical brain imaging applications.
Researcher Affiliation Academia Yuliang Xu Department of Statistical Science Duke University EMAIL Jian Kang Department of Biostatistics University of Michigan EMAIL
Pseudocode No The paper describes algorithms like CAVI and SSVI, providing update equations and general descriptions (e.g., 'Eq(t) {βj} = Eq(t 1) {βj} + st n/ns Eq(t 1) log X i I p(Yi, Mi, Xi | θ) + Eq(t 1) log π(βj)'), but it does not present them in a structured pseudocode or algorithm block format.
Open Source Code Yes 1STCAR is available as an R package on Github https://github.com/yuliangxu/STCAR
Open Datasets Yes We further validate our approach by applying the ST-CAR to working memory f MRI data from the Adolescent Brain Cognitive Development (ABCD) study
Dataset Splits Yes The entire data set is split into 70% training data and 30% testing data.
Hardware Specification No The paper discusses computation time for various methods and models but does not provide specific details about the hardware used, such as CPU or GPU models, memory, or cloud instance types.
Software Dependencies No The paper mentions several software packages and libraries like glmnet, BIMA, T-Lo Ho, Rcpp Armadillo, and RANN, along with their respective citations. However, it does not provide specific version numbers for any of these software dependencies.
Experiment Setup Yes The thresholding parameter ν is set to be the largest marginal value in β estimated from ridge regression... The decay rate of σ2 β is set to be 0.5(1 + t) 0.7 where t represents the number of iterations... For the neighboring matrix B, we set the number of neighbors as 8, and the correlation parameter ρ is set to be 0.9. The variance parameter σµ is fixed at 1 for the CAVI update. In the ST-CAR method implemented using the SSVI algorithm... We use a step of 10 4 and a subsample of 100 for the SGD optimization. The decay rate function for σ2 α is (1 + t) 0.4.