LaplaceErrorDistribution = class LaplaceErrorDistribution(ScaledErrorDistribution) |
|
LaplaceErrorDistribution(scale=1.0, limits=None, copy=None)
To calculate a Laplace likelihood.
For one residual, x, it holds
f( x ) = 1 / ( 2 s ) exp( - |x| / s )
where s is the scale.
s is a hyperparameter, which might be estimated from the data.
The variance of this function is σ^2 = 2 s ^ 2.
See: toSigma()
The function is mostly used to calculate the likelihood L over N
residuals, or easier using log likelihood, logL.
logL = log( N / ( 2 s ) ) - ∑( |x| / s )
Using weights this becomes:
logL = log( ∑( w ) / ( 2 s ) ) - ∑( w |x| / s )
Using this error distribution results in median-like solutions.
Author Do Kester. |
|
- Method resolution order:
- LaplaceErrorDistribution
- ScaledErrorDistribution
- ErrorDistribution
- builtins.object
Constructor:
- LaplaceErrorDistribution( scale=1.0, limits=None, copy=None )
- Constructor of Laplace Distribution.
Parameters
----------
scale : float
noise scale
limits : None or list of 2 floats [low,high]
None : no limits implying fixed scale
low low limit on scale (needs to be >0)
high high limit on scale
when limits are set, the scale is *not* fixed.
copy : LaplaceErrorDistribution
distribution to be copied.
Methods defined here:
- acceptWeight()
- True if the distribution accepts weights.
Always true for this distribution.
- copy()
- Return copy of this.
- getScale( problem, allpars=None )
- Return the noise scale
Parameters
----------
problem : Problem
to be solved
allpars : array_like
None take parameters from problem.model
list of all parameters in the problem
- getSumRes( problem, allpars=None, scale=1)
- Return the sum of the absolute values of the residuals.
sum ( | res | )
Parameters
----------
problem : Problem
to be solved
allpars : array_like
None take parameters from problem.model
list of all parameters in the problem
scale : float or array_like
scale of residuals (from accuracies or noisescale of errdis)
- logLdata( problem, allpars, mockdata=None )
- Return the log( likelihood ) for each residual
logL = sum( logLdata )
Parameters
----------
problem : Problem
to be solved
allpars : array_like
list of all parameters in the problem
mockdata : array_like
as calculated by the model
- logLikelihood_alt( problem, allpars )
- Return the log( likelihood ) for a Gaussian distribution.
Parameters
----------
problem : Problem
to be solved
allpars : array_like
parameters of the problem
- nextPartialData( problem, allpars, fitIndex, mockdata=None )
- Return the partial derivative of elements of the log( likelihood )
to the parameters.
dL/ds is not implemented for problems with accuracy
Parameters
----------
problem : Problem
to be solved
allpars : array_like
list of all parameters in the problem
fitIndex : array_like
indices of parameters to be fitted
mockdata : array_like
as calculated by the model
- partialLogL_alt( problem, allpars, fitIndex )
- Return the partial derivative of log( likelihood ) to the parameters.
dL/ds is not implemented for problems with accuracy
Parameters
----------
problem : Problem
to be solved
allpars : array_like
list of all parameters in the problem
fitIndex : array_like
indices of parameters to be fitted
- toSigma( scale )
- Return sigma, the squareroot of the variance.
Parameter
--------
scale : float
the scale of this Laplace distribution.
Methods inherited from ScaledErrorDistribution:
Methods inherited from ErrorDistribution:
- domain2Unit( dval, ks )
- getChisq( problem, allpars=None )
- getGaussianScale( problem, allpars=None )
- getResiduals( problem, allpars=None )
- hyparname( k )
- isBound()
- keepFixed( fixed=None )
- logCLhood( problem, allpars )
- logLhood( problem, allpars )
- numPartialLogL( problem, allpars, fitIndex )
- partialLogL( problem, allpars, fitIndex )
- setPriors( priors )
- setResult()
- unit2Domain( uval, ks )
- updateLogL( problem, allpars, parval=None )
|
|