Then what about 9 AM Pacific Time on Saturday morning? That will keep it as close to the original S4TF team’s meeting schedule as possible, still being sort of symbolic.
@aj_ortiz +1 on what you said regarding the similarity between numerical computing and ML. I want to add that Swift AutoDiff has allowed things like “Deep Physics” at PassiveLogic that blur the line between numerical computing and ML. Also, both TF and Torch have both linear algebra and machine learning modules, and even S4TF has some linear algebra operators. So the two fields are very intertwined in practice.
Also since ML is just gradient descent optimization, it could be considered a subset of numerical computing. And our community might already be so small that splintering it up into two work groups might destroy our effectiveness at doing things.