SGP sometimes creates too-coarse approximation of the input data. Reading through the documentation, I cannot find a way to trade runtime for accuracy. Maybe increase n_start or n_inducing, or switch to the VFE method?
Willing to submit a PR once the situation is clarified!
SGP sometimes creates too-coarse approximation of the input data. Reading through the documentation, I cannot find a way to trade runtime for accuracy. Maybe increase
n_startorn_inducing, or switch to theVFEmethod?Willing to submit a PR once the situation is clarified!