Block Sparse Bayesian Learning: A Diversified Scheme

Yanhao Zhang, Zhihan Zhu, Yong Xia
2/3/2026

No official code found. Best implementation: YanhaoZhang1/DivSBL (19 stars, last commit 2024-10-28).

Abstract

This paper introduces a novel prior called Diversified Block Sparse Prior to characterize the widespread block sparsity phenomenon in real-world data. By allowing diversification on intra-block variance and inter-block correlation matrices, we effectively address the sensitivity issue of existing block sparse learning methods to pre-defined block information, which enables adaptive block estimation while mitigating the risk of overfitting. Based on this, a diversified block sparse Bayesian learning method (DivSBL) is proposed, utilizing EM algorithm and dual ascent method for hyperparameter estimation. Moreover, we establish the global and local optimality theory of our model. Experiments validate the advantages of DivSBL over existing algorithms.

View on arXivView PDF

Code Implementations(1)

YanhaoZhang1/DivSBLCommunity100%

The main function for Sparse Bayesian Learning: A Diversified Scheme (Accepted to NeurIPS 2024)

191C, CSSFeb 4, 20241 years ago

Similar Papers

SPP-SBL: Space-Power Prior Sparse Bayesian Learning for Block Sparse Recovery

80% match1 repo82025

An Efficient Training Algorithm for Models with Block-wise Sparsity

80% match2025

R-Sparse: Rank-Aware Activation Sparsity for Efficient LLM Inference

79% match1 repo212026

Sparsity via Hyperpriors: A Theoretical and Algorithmic Study under Empirical Bayes Framework

79% match2025

Meta-Sparsity: Learning Optimal Sparse Structures in Multi-task Networks through Meta-learning

79% match2025

SUBP: Soft Uniform Block Pruning for 1xN Sparse CNNs Multithreading Acceleration

78% match1 repo222026

Cite this paper

@article{zhang2026block,
  title  = {Block Sparse Bayesian Learning: A Diversified Scheme},
  author = {Yanhao Zhang and Zhihan Zhu and Yong Xia},
  year   = {2026},
  eprint = {2402.04646},
  archivePrefix = {arXiv},
  doi    = {10.52202/079017-4130},
  url    = {https://arxiv.org/abs/2402.04646},
  journal = {NEURIPS 2024 2024}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2402.04646.svg)](https://codeofpaper.com/paper/2402.04646)

Discussion