It seems like the smoothing splines can heavily distort the geometry of certain trajectories, resulting in poor kinematic variable guesses that the least-squares minimizer cannot recover from but the LBFGS minimizer can. Needs some investigation into this.
It seems like the smoothing splines can heavily distort the geometry of certain trajectories, resulting in poor kinematic variable guesses that the least-squares minimizer cannot recover from but the LBFGS minimizer can. Needs some investigation into this.