CISPA
Browse

File(s) not publicly available

On (1 + ϵ)-Approximate Block Sparse Recovery

conference contribution
posted on 2023-11-29, 18:20 authored by Baris Can EsmerBaris Can Esmer, Vasileios Nakos
Learning approximately block sparse vectors using a small number of linear measurements is a standard task in the sparse recovery/compressed sensing literature. Schemes achieving a constant factor approximation are long known, e.g. using model-based RIP. We give a new scheme achieving (1+\epsilon) approximation, which runs in near linear time in the length of the vector and is likely to be optimal up to constant factors. As an intriguing side result, we obtain the simplest known scheme measurement-optimal \ell_2/\ell_2 sparse recovery scheme recorded in the literature. The main component of our algorithm is a subtle variant of the classic CountSketch data structure where the random signs are substituted by gaussians and the number of repetitions (rows) is tuned to smaller than usual.

History

Preferred Citation

Baris Esmer and Vasileios Nakos. On (1 + ϵ)-Approximate Block Sparse Recovery. In: IEEE International Symposium on Information Theory (ISIT). 2022.

Primary Research Area

  • Algorithmic Foundations and Cryptography

Name of Conference

IEEE International Symposium on Information Theory (ISIT)

Legacy Posted Date

2022-05-05

Open Access Type

  • Unknown

BibTeX

@inproceedings{cispa_all_3648, title = "On (1 + ϵ)-Approximate Block Sparse Recovery", author = "Esmer, Baris Can and Nakos, Vasileios", booktitle="{IEEE International Symposium on Information Theory (ISIT)}", year="2022", }

Usage metrics

    Categories

    No categories selected

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC