pystra.active_learning.pc_kriging.PCKrigingFitResult

Use it: User guide · Worked example · Theory

class PCKrigingFitResult(trend, coefficients, length_scale, process_variance, optimized, objective)

Bases: object

Snapshot conditional on the selected trend and fitted kernel.

trend records the preliminary sparse PCE; coefficients are the subsequent GLS trend coefficients in physical response units. Length scales follow input column order. Predictive variance does not integrate uncertainty in basis selection or kernel hyperparameters. optimized records whether the length scales were fitted. objective is the profiled likelihood criterion per observation for the scaled response, with additive constants omitted; it is not a reliability error metric.

Methods