BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Improved data covariance estimation techniques applied to lattice
QCD
DTSTART;VALUE=DATE-TIME:20170622T163000Z
DTEND;VALUE=DATE-TIME:20170622T165000Z
DTSTAMP;VALUE=DATE-TIME:20200220T051115Z
UID:indico-contribution-386@cern.ch
DESCRIPTION:Speakers: Dr. SIMONE\, james (fermilab)\nQuantities in lattice
QCD are estimated from statistical fits to\nlattice correlation functions
. For example\, a meson mass may be\nestimated from a fit to a two-point
function\, $C(t)$\, computed at $p$\ntimeslices\, where each measurement i
s averaged over a sample of $n$\,\nstatistically independent gauge configu
rations. Correlations among\n$C(t)$ at nearby time slices means that the
statistical procedure must\nincludes an estimate of the rank $p$ the covar
iance matrix. Most\noften\, the desired inverse of the population covaria
nce matrix is\nsimply estimated to be equal to the inverse of the sample c
ovariance\nmatrix. The sample covariance\, however\, is known to be a poor
estimate\nof the population covariance when $p / n$ is near one. In fact\
, the\nsample covariance matrix will have one or more zero eigenvalues whe
n\n$p / n \\ge 1$. Cases where $p / n \\sim 1$ are encountered in\npracti
ce\, where the sample covariance matrix is found to have a number\n"small"
eigenvalues. In this talk we discuss the application of\nlinear and nonl
inear "shrinkage" techniques to the estimation of\npopulation covariance.
For the nonlinear shrinkage technique we\nconsider\, the estimated covaria
nce matrix approaches the population\ncovariance matrix in the limit $p\,
n \\to \\infty$\, and $p / n =\n\\mathrm{constant}$.\n\nhttps://makondo.ug
r.es/event/0/session/102/contribution/386
LOCATION: Seminarios 8
URL:https://makondo.ugr.es/event/0/session/102/contribution/386
END:VEVENT
END:VCALENDAR