Sparse Hyperspectral Band Selection Based on Expectation Maximization

Likun Gao, Hai-Miao Hu, Xinhui Xue, Haowen Zheng

Keywords: Hyperspectral Band Selection, Sparse Learning, Feature Selection, Hyperspectral Image Classification, EM Algorithm

TL;DR: This paper proposes a sparse learning method for spectral band selection based on the EM algorithm, which effectively characterizes the relationships between spectral bands to achieve state-of-the-art performance.

Abstract: Band selection is crucial in spectral imaging, as it involves choosing the most relevant bands from large hyperspectral datasets to retain essential information while reducing the burden of data transmission and analysis. Addressing this need, we introduce a novel method for band selection that utilizes an Expectation Maximization algorithm to facilitate selection through the sparsification of spectral band importance. Our method enhances sparsity effects and effectively delineates the relationships between spectral bands during the sparsification process. Supported by thorough theoretical analysis and experimental validation on public datasets, our approach has proven to be both robust and practical. Compared to other sparsification methods, it not only excels in achieving significant sparsity effects but also demonstrates marked advantages in illustrating inter-band relationships. Our method delivers outstanding performance in band selection tasks and holds potential for broader applications in other sparsity-oriented contexts in the future.

Primary Area: learning theory

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 5562