Inference¶

Fitting a model to data, and sampling its posterior.

Optimizer¶

`FitSummary`	Basic fit information.
`Optimizer`	Fit any pytree of Parameters with a minimiser or least-squares solver.

`parameter_correlation`(fitted, residuals_fn, *)	Full covariance and correlation of the free parameters at fitted.
`parameter_errors`(fitted, residuals_fn, *[, ...])	1-sigma errors on every trainable Parameter, from the Gauss-Newton covariance.
`rescale_from_errors`(fitted, errs)	Rescale every trainable Parameter using its error estimate.

How a fit's parameters trade off against one another.

A profile-likelihood scan: chi-squared over a grid, nuisances refitted.

Joint fit over several targets with shared instruments.

record_fit(opt, model[, probes, max_steps, ...])

Step opt over model, recording the loss and the probes.

The path a fit took, sampled once per recorded solver step.

`scene_model`(model, data, priors, likelihood)	A NumPyro model sampling model's unfrozen parameters.
`free_parameters`(model)	Shape of every unfrozen parameter: the sites `scene_model()` samples.
`init_values`(model)	Free parameter values, keyed to `scene_model()`'s sites.