linearization
The linearisation of a fit at its MAP, and the linearised posterior covariance.
A linearisation is a Gauss–Newton approximation of a fit: the Jacobian
\(J = -\partial r / \partial \theta\) of the event-space residual with respect to the
model’s free parameters, with the discrepancy’s hyperparameters held fixed, and the
Fisher matrix \(F = \sum_b J_b^\top \Sigma_{D,b}^{-1} J_b\) with
\(\Sigma_D = K + \Sigma_n\) per event block. Build one with
pmrf.objectives.evaluators.MarginalLogLikelihood.linearize(), and combine one or more with
the prior in posterior_covariance().
Classes
|
The linearisation of one fit at a model. |
Functions
|
The linearised posterior covariance of a model's free parameters. |