Programmer Guide/Command Reference/EVAL/density: Difference between revisions

From STX Wiki
Jump to navigationJump to search
(Created page with '{{DISPLAYTITLE:{{SUBPAGENAME}}}} Evaluates the log-likelihood of (a) feature vector(s) for a mixture model specified in the parameters meantable, sigmatable and wtable (weights).…')
 
No edit summary
 
(One intermediate revision by the same user not shown)
Line 6: Line 6:
:;<var>sigmatable</var>:covariance vectors (i.e. diagonal of covariance matrices) of mixture model (N<sub>mixtures</sub> x N<sub>dimensions</sub>)
:;<var>sigmatable</var>:covariance vectors (i.e. diagonal of covariance matrices) of mixture model (N<sub>mixtures</sub> x N<sub>dimensions</sub>)
:;<var>wtable</var>:vector of mixture weights
:;<var>wtable</var>:vector of mixture weights
:;<var>accurate</var>
:;<var>accurate</var>:
::*if '''1''', the accurate log-likelihood estimation is used
::*if '''1''', the accurate log-likelihood estimation is used
::*otherwise use the Intel IPP fast log. add method
::*otherwise use the Intel IPP fast log. add method
:;<var>divbylength</var>
:;<var>divbylength</var>:
::*if '''1''', divide overall log-likelihood by number of observations
::*if '''1''', divide overall log-likelihood by number of observations
::*otherwise not
::*otherwise not
:;<var>framell</var>
:;<var>framell</var>:
::*if '''1''', log-likelihoods are calculated for each individual frame
::*if '''1''', log-likelihoods are calculated for each individual frame
::*otherwise they are summed up
::*otherwise they are summed up
:;<var>lowerquantile</var>if a value is provided, likelihood values below this value are discarded
:;<var>lowerquantile</var>:if a value is provided, likelihood values below this value are discarded
:;<var>upperquantile</var>if a value is provided, likelihood values above this value are discarded
:;<var>upperquantile</var>:if a value is provided, likelihood values above this value are discarded
;Result: Joint log-likelihood, for each frame (if ''framell''=0) or summed up (if ''framell''=1) and divided by number of observations (if ''divbylength''=1).
;Result: Joint log-likelihood, for each frame (if ''framell''=0) or summed up (if ''framell''=1) and divided by number of observations (if ''divbylength''=1).
;See also: [[../haclust|haclust]], [[../modclust|modclust]], [[../em|em]], [[../svd|svd]]
;See also: [[../haclust|haclust]], [[../modclust|modclust]], [[../em|em]], [[../svd|svd]]


[[../#Functions|<function list>]]
[[../#Functions|<function list>]]

Latest revision as of 07:58, 21 April 2011

Evaluates the log-likelihood of (a) feature vector(s) for a mixture model specified in the parameters meantable, sigmatable and wtable (weights).

Usage
density(x, meantable, sigmatable, wtable, accurate, divbylength, framell {, lowerquantile, upperquantile})
x
input feature vectors
meantable
mean vectors of mixture model (Nmixtures x Ndimensions)
sigmatable
covariance vectors (i.e. diagonal of covariance matrices) of mixture model (Nmixtures x Ndimensions)
wtable
vector of mixture weights
accurate
  • if 1, the accurate log-likelihood estimation is used
  • otherwise use the Intel IPP fast log. add method
divbylength
  • if 1, divide overall log-likelihood by number of observations
  • otherwise not
framell
  • if 1, log-likelihoods are calculated for each individual frame
  • otherwise they are summed up
lowerquantile
if a value is provided, likelihood values below this value are discarded
upperquantile
if a value is provided, likelihood values above this value are discarded
Result
Joint log-likelihood, for each frame (if framell=0) or summed up (if framell=1) and divided by number of observations (if divbylength=1).
See also
haclust, modclust, em, svd

<function list>

Navigation menu

Personal tools