List-decodable covariance estimation
List-decodable covariance estimation
复制标题
列表可解码协方差估计
DOI:
10.1145/3519935.3520006
复制
发表时间:
2022
期刊:
影响因子:
--
通讯作者:
Kothari, Pravesh K.
中科院分区:
文献类型:
--
作者:
Ivkov, Misha;Kothari, Pravesh K.
We give the first polynomial time algorithm forlist-decodable covariance estimation. For any α > 0, our algorithm takes input a sampleY⊆dof sizen≥dpoly(1/α)obtained by adversarially corrupting an (1−α)npoints in an i.i.d. sampleXof sizenfrom the Gaussian distribution with unknown mean µ*and covariance Σ*. Innpoly(1/α)time, it outputs a constant-size list ofk=k(α)= (1/α)poly(1/α)candidate parameters that, with high probability, contains a (µ,Σ) such that the total variation distanceTV(N(µ*,Σ*),N(µ,Σ))<1−Oα(1). This is a statistically strongest notion of distance and implies multiplicative spectral and relative Frobenius distance approximation with dimension independent error. Our algorithm works more generally for any distributionDthat possesses low-degree sum-of-squares certificates of two natural analytic properties: 1) anti-concentration of one-dimensional marginals and 2) hypercontractivity of degree 2 polynomials.Prior to our work, the only known results for estimating covariance in the list-decodable setting were for the special cases of list-decodable linear regression and subspace recovery [Karmalkar-Klivans-Kothari 2019, Bakshi-Kothari 2020, Raghavendra-Yau’19, 20]. Even for these special cases, the known error guarantees are weak and in particular, the algorithms need super-polynomial time for any sub-constant (in dimensiond) target error in natural norms. Our result, as a corollary, yields the first polynomial timeexactalgorithm for list-decodable linear regression and subspace recovery that, in particular, obtain 2−(d)error in polynomial-time in the underlying dimension.
登录
查看更多内容
DOI:
--
发表时间:
2019
期刊:
Advances in neural information processing systems
影响因子:
--
作者:
Karmalkar, Sushrut;Klivans, Adam;Kothari, Pravesh
通讯作者:
Kothari, Pravesh
DOI:
--
发表时间:
2017
期刊:
arXiv.org
影响因子:
--
作者:
Pravesh Kothari;David Steurer
通讯作者:
David Steurer
DOI:
--
发表时间:
2020
期刊:
arXiv.org
影响因子:
--
作者:
Ainesh Bakshi;Pravesh Kothari
通讯作者:
Pravesh Kothari
DOI:
10.1145/3188745.3188758
发表时间:
2017-11
期刊:
Proceedings of the 50th Annual ACM SIGACT Symposium on Theory of Computing
影响因子:
--
作者:
Ilias Diakonikolas;D. Kane;Alistair Stewart
通讯作者:
Ilias Diakonikolas;D. Kane;Alistair Stewart
DOI:
10.1145/3357713.3384329
发表时间:
2019-12
期刊:
Proceedings of the 52nd Annual ACM SIGACT Symposium on Theory of Computing
影响因子:
--
作者:
Yeshwanth Cherapanamjeri;Samuel B. Hopkins;Tarun Kathuria;P. Raghavendra;Nilesh Tripuraneni
通讯作者:
Yeshwanth Cherapanamjeri;Samuel B. Hopkins;Tarun Kathuria;P. Raghavendra;Nilesh Tripuraneni