ICML 2015poster121 citations
A Nearly-Linear Time Framework for Graph-Structured Sparsity
Chinmay Hegde, Piotr Indyk, Ludwig Schmidt
Abstract
We introduce a framework for sparsity structures defined via graphs. Our approach is flexible and generalizes several previously studied sparsity models. Moreover, we provide efficient projection algorithms for our sparsity model that run in nearly-linear time. In the context of sparse recovery, we show that our framework achieves an information-theoretically optimal sample complexity for a wide range of parameters. We complement our theoretical analysis with experiments demonstrating that our algorithms improve on prior work also in practice.
BibTeX
@InProceedings{pmlr-v37-hegde15,
title = {A Nearly-Linear Time Framework for Graph-Structured Sparsity},
author = {Hegde, Chinmay and Indyk, Piotr and Schmidt, Ludwig},
booktitle = {Proceedings of the 32nd International Conference on Machine Learning},
pages = {928--937},
year = {2015},
editor = {Bach, Francis and Blei, David},
volume = {37},
series = {Proceedings of Machine Learning Research},
address = {Lille, France},
month = {07--09 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v37/hegde15.pdf},
url = {https://proceedings.mlr.press/v37/hegde15.html},
abstract = {We introduce a framework for sparsity structures defined via graphs. Our approach is flexible and generalizes several previously studied sparsity models. Moreover, we provide efficient projection algorithms for our sparsity model that run in nearly-linear time. In the context of sparse recovery, we show that our framework achieves an information-theoretically optimal sample complexity for a wide range of parameters. We complement our theoretical analysis with experiments demonstrating that our algorithms improve on prior work also in practice.}
}