Number of relevant directions in Principal Component Analysis and Wishart random matrices

arXiv1112.5391MaRDI QIDQ6229869FDOQ6229869

Satya N. Majumdar, P. Vivo

Publication date: 22 December 2011

Abstract: We compute analytically, for large $N$, the probability $mathcalP(N_{+},N)$ that a $NimesN$ Wishart random matrix has $N_{+}$ eigenvalues exceeding a threshold $Nzeta$, including its large deviation tails. This probability plays a benchmark role when performing the Principal Component Analysis of a large empirical dataset. We find that $$, where $$ is the Dyson index of the ensemble and $ps{i}_{z}eta(kappa)$ is a rate function that we compute explicitly in the full range $0leqkappaleq1$ and for any $zeta$. The rate function $ps{i}_{z}eta(kappa)$ displays a quadratic behavior modulated by a logarithmic singularity close to its minimum $kapp{a}^{s}tar(zeta)$. This is shown to be a consequence of a phase transition in an associated Coulomb gas problem. The variance $Delta(N)$ of the number of relevant components is also shown to grow universally (independent of $zeta)$ as $$ for large $N$.

This page was built for publication: Number of relevant directions in Principal Component Analysis and Wishart random matrices

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6229869)