This is a follow-up to #161: For very large analysis that are built incrementally, all algorithms should be able to build on previous results and only extend it for new epochs. This requires both a suitable invalidation mechanism (easy) and an adoption of algorithms to restrict their data scope. The latter also needs an API that will require some thought.
This is a follow-up to #161: For very large analysis that are built incrementally, all algorithms should be able to build on previous results and only extend it for new epochs. This requires both a suitable invalidation mechanism (easy) and an adoption of algorithms to restrict their data scope. The latter also needs an API that will require some thought.