Initialization is fundamental for the 6D version due to radial velocity uncertainties and possible outliers.
Unfortunately in the init_nuts function of pymc convergence tolerance is hard-coded. If the sampler fails with the derivative of the RV is zero it meas the initialization is not good enough and we will need to modify those tolerances.
Initialization is fundamental for the 6D version due to radial velocity uncertainties and possible outliers. Unfortunately in the init_nuts function of pymc convergence tolerance is hard-coded. If the sampler fails with the derivative of the RV is zero it meas the initialization is not good enough and we will need to modify those tolerances.
Solutions: