Monotonicity conjecture for the number of spectral components
Let be the limiting spectral distribution of the noisy sample covariance matrix at noise power , and let denote the number of connected components of its support. Under the standing assumptions of the paper, for all . Monotonicity conjecture. is non-increasing in . This would imply that once the separation between the information and noise eigenvalues disappears, it does not reappear as the noise power increases; the paper gives simulations and notes analogous results for other random matrix models, but does not establish the claim.
References
- Mario Diaz, Shahab Asoodeh, Fady Alajaji, Tamás Linder, Serban Belinschi and James Mingo, “On the Noise-Information Separation of a Private Principal Component Analysis Scheme”, arXiv:1801.03553 (2018).
Progress summary
Never refreshed
Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.
Solutions 0
No solutions have been posted yet.