The canonical way to put a prior on functions is with a Gaussian process. It’s essentially a continuous-time limit of the GaussianRandomWalk and the PyMC3 API for it is really nice. I haven’t used the SDE functionality yet, so no comment on that.
The canonical way to put a prior on functions is with a Gaussian process. It’s essentially a continuous-time limit of the GaussianRandomWalk and the PyMC3 API for it is really nice. I haven’t used the SDE functionality yet, so no comment on that.