rcrpy.Priors

class rcrpy.Priors(prior_type=None, p=None, gaussian_params=None, param_bounds=None)[source]

Bases: object

Probabilistic priors on model parameters.

Mirrors the C++ Priors class (cpp/src/FunctionalForm.h:22). Four modes:

  • GAUSSIAN: gaussian_params[i] = (mu_i, sigma_i). NaN values mean no prior on that param. Implemented as quadratic penalty residuals (params[i] - mu_i) / sigma_i appended to the least-squares residual.

  • CONSTRAINED: param_bounds[i] = (lb_i, ub_i). NaN means unbounded on that side. Wired into scipy.optimize.least_squares as its bounds= argument.

  • MIXED: both Gaussian and Constrained simultaneously.

  • CUSTOM: user supplies p(params) -> ndarray which is appended to the residual vector verbatim (allows arbitrary log-posterior shape).

Construct with the appropriate mode + data; the regression() routine in FunctionalForm reads prior_type, gaussian_params, param_bounds, and p as needed.

Parameters:

  • prior_type (PriorType | None)

  • p (Callable[[np.ndarray], np.ndarray] | None)

  • gaussian_params (Sequence[Sequence[float]] | None)

  • param_bounds (Sequence[Sequence[float]] | None)

__init__(prior_type=None, p=None, gaussian_params=None, param_bounds=None)[source]
Parameters:

Methods

__init__([prior_type, p, gaussian_params, ...])

gaussian_residuals(params)

For active Gaussian priors, return (params[i] - mu) / sigma for each param with a defined (non-NaN) prior.

scipy_bounds(n_params)

Convert param_bounds into the (low, high) arrays scipy expects (with np.inf substituted for NaN — i.e. unbounded).
gaussian_residuals(params)[source]

For active Gaussian priors, return (params[i] - mu) / sigma for each param with a defined (non-NaN) prior. Empty array if no Gaussian priors active.

Parameters:

params (ndarray)

Return type:

ndarray

scipy_bounds(n_params)[source]

Convert param_bounds into the (low, high) arrays scipy expects (with np.inf substituted for NaN — i.e. unbounded). For Gaussian- or no-prior cases, returns the trivial all-unbounded pair.

Parameters:

n_params (int)

Return type:

tuple[ndarray, ndarray]