pint.fitter.ModelState
- class pint.fitter.ModelState(fitter, model)[source]
Bases:
object
Record a model state and cache calculations
This class keeps track of a particular model state and all the associated matrices - design matrices, singular value decompositions, what have you - that are needed to compute a step and evaluate the quality of the fit.
These objects should be regarded as immutable but lazily evaluated.
Methods
predicted_chi2
(step, lambda_)Predict the chi2 after taking a step based on the linear approximation
take_step
(step, lambda_)Return a new state moved by lambda_*step.
take_step_model
(step[, lambda_])Make a new model reflecting the new parameters.
Attributes
chi2
covariance_matrix
parameter_covariance_matrix
resids
step