Why not run the query search and model update in parallel? Currently they're sequential.
This will eliminate the need to balance between time spent searching and time spent updating, at least for a moderately small number of adaptive algorithms. It should be fine to run in parallel as long as there are Dask workers/cores available.
35 is relevant because executing in parallel will require that some basic inter-process communication.
Why not run the query search and model update in parallel? Currently they're sequential.
This will eliminate the need to balance between time spent searching and time spent updating, at least for a moderately small number of adaptive algorithms. It should be fine to run in parallel as long as there are Dask workers/cores available.
35 is relevant because executing in parallel will require that some basic inter-process communication.