컨텐츠 시작
학술대회/행사
초록검색
| 제출번호(No.) | 0253 |
|---|---|
| 분류(Section) | Special Session |
| 분과(Session) | Structured nonparametric and high-dimensional statistics (SS-15) |
| 영문제목 (Title(Eng.)) |
Large ball probability and inference for spectral projectors |
| 저자(Author(s)) |
Vladimir Spokoiny1 WIAS and HU Berlin1 |
| 초록본문(Abstract) | Let \( X_{1}, \ldots ,X_{n} \) be i.i.d. sample in \( R^{p} \) with zero mean and the covariance matrix \( \Sigma \). We consider the problem of confidence estimation of the projectors on an eigenspace of \( \Sigma \). This paper offers two procedures: one is based on the resampling technique; the other one is Bayesian and uses Bayesian calculus from the conjugated Wishart prior. Accuracy of both methods is evaluated with sharp error bounds. The study heavily uses recent results on “large ball probability” for Gaussian measures in a Hilbert space. |
| 분류기호 (MSC number(s)) |
62G05, 62E17 |
| 키워드(Keyword(s)) | Gaussian comparison, Gaussian anti-concentration inequalities, dimension free bounds, Schatten norm, high-dimensional inference, high-dimensional CLT |
| 강연 형태 (Language of Session (Talk)) |
English |