bayespecon.logdet.flow_logdet_numpy¶

bayespecon.logdet.flow_logdet_numpy(rho_d, rho_o, rho_w, poly_a, poly_b, poly_c, poly_coeffs, miter_a, miter_b, miter_c, miter_coeffs, miter, titer=800)[source]¶

Vectorised numpy port of flow_logdet_pytensor().

Evaluates \(\log|I_N - \rho_d W_d - \rho_o W_o - \rho_w W_w|\) for arrays of posterior draws (rho_d, rho_o, rho_w). Used by FlowModel-derived models to attach a complete pointwise log-likelihood to the InferenceData after sampling (the pm.Potential("jacobian", ...) Jacobian term is not captured by PyMC’s automatic log-likelihood accounting).

Parameters:¶

rho_d : array-like, shape (G,) or scalar¶

Posterior draws for the three spatial parameters.

rho_o : array-like, shape (G,) or scalar¶

Posterior draws for the three spatial parameters.

rho_w : array-like, shape (G,) or scalar¶

Posterior draws for the three spatial parameters.

poly_a¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

poly_b¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

poly_c¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

poly_coeffs¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

miter_a¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

miter_b¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

miter_c¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

miter_coeffs¶

Precomputed exponent arrays and coefficients from _flow_logdet_poly_coeffs().

miter : int¶

Highest polynomial order included in the exact series.

titer : int, default 800¶

Highest order included in the geometric tail approximation.

Returns:¶

Log-determinant evaluated at each draw.

Return type:¶

np.ndarray, shape (G,)