nifty8.re.minisanity module#

class ChiSqStats(mean, reduced_chisq, ndof)[source]#

Bases: NamedTuple

minisanity(position_or_samples, func=None, *, map='lmap')[source]#: Wrapper for reduced_residual_stats to retrieve the reduced chi-squared and a pretty-printable string of the statistics.

reduced_residual_stats(position_or_samples, func=None, *, map='lmap')[source]#

Computes the average, reduced chi-squared, and number of parameters as a summary statistics for a given input.

Parameters:#

position_or_samples: tree-like or Samples: Input values to compute reduces chi-sq statistics. The statistics is computed for each leaf of the pytree, i.E. only array-like leafs are square averaged. If positin_or_samples is a Sample object, the chi-sq statistics is computed for each sample, and the sample mean and standard deviation of the statistics is returned.
func: Callable (optional): Function to apply to position_or_samples before computing the chi-sq statistics for. If provided, the statistics is computed for func(x) instead of x where x is either primals or a sample.

stats: tree-like: Pytree of tuple containing the mean, reduced chi-squared, and number of parameters for each leaf of the input tree. For the mean and reduched chi-sq, a numpy array with the sample mean and sample std is returned. If samples is None, the second entry of this array is always zero.