You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Expecting users to convert and pass the appropriate version of the graph such as transpose or symmetric for different analytics algorithms is very inconvenient.
We can amortize the cost of conversion by storing the transformed RDG.
The text was updated successfully, but these errors were encountered:
We discussed tagging RDG versions as having specific properties and maybe even being derived from specific other RDGs. It should be integrated into the existing algorithms so they can avoid the documentation only assumptions of symmetry and the like. I think the result of the discussion was to push the idea back a bit and this may still not be the time to implement it fully, but I don't want to get it lost.
We now have the in-memory operations to do this (d5dc5fb). However the algorithms don't use them, and don't have a way to know if they should use them. We will need to extend the graph metadata to include the needed information.
Expecting users to convert and pass the appropriate version of the graph such as transpose or symmetric for different analytics algorithms is very inconvenient.
We can amortize the cost of conversion by storing the transformed RDG.
The text was updated successfully, but these errors were encountered: