CRISPy's efficiency can be significantly improved if only the data points within a few smoothing lengths of a walker are used to compute SCMS for each walker. It'd be good to provide ways to turn this option off just in case such an approximate is bad for special use cases.
CRISPy's efficiency can be significantly improved if only the data points within a few smoothing lengths of a walker are used to compute SCMS for each walker. It'd be good to provide ways to turn this option off just in case such an approximate is bad for special use cases.