ML Anthology

Subspace Detection: A Robust Statistics Formulation

Sanjoy Dasgupta
COLT 2003 pp. 734
doi:10.1007/978-3-540-45167-9_55 /colt/2003/dasgupta2003colt-subspace/

Abstract

If data in ℝ^ d actually lie in a linear subspace, then principal component analysis (PCA) will find this subspace. If the data are corrupted by benign (eg. independent Gaussian) noise, then approximation bounds can quite easily be shown for the solution returned by PCA. What if the noise is malicious?

PDF COLT Semantic Scholar

Cite

Text

Dasgupta. "Subspace Detection: A Robust Statistics Formulation." Annual Conference on Computational Learning Theory, 2003. doi:10.1007/978-3-540-45167-9_55

Markdown

[Dasgupta. "Subspace Detection: A Robust Statistics Formulation." Annual Conference on Computational Learning Theory, 2003.](https://mlanthology.org/colt/2003/dasgupta2003colt-subspace/) doi:10.1007/978-3-540-45167-9_55

BibTeX

@inproceedings{dasgupta2003colt-subspace,
  title     = {{Subspace Detection: A Robust Statistics Formulation}},
  author    = {Dasgupta, Sanjoy},
  booktitle = {Annual Conference on Computational Learning Theory},
  year      = {2003},
  pages     = {734},
  doi       = {10.1007/978-3-540-45167-9_55},
  url       = {https://mlanthology.org/colt/2003/dasgupta2003colt-subspace/}
}