More heuristics to cut the design space down

drbenvincent / darc_toolbox

Run adaptive decision making experiments

MIT License

16 stars 2 forks source link

More heuristics to cut the design space down #27

Closed drbenvincent closed 5 years ago

drbenvincent commented 5 years ago

Basic heuristics are done in #11. Need to implement the more advanced ones

[x] Implement the heuristic_order approach: Done with choose_one_along_design_dimension()
[x] allow design dimension to be defined from a higher-level point

drbenvincent commented 5 years ago

choose_one_along_design_dimension() seems to be working ok for regular delay discounting, but causing crashes (no designs) for magnitude effect designs (see #28)

I think at this point we need to bring in the design + model integration tests (see #33) because we are getting frustrating errors coming up a lot with magnitude effect design spaces

drbenvincent commented 5 years ago

At this point we've got good a good heuristic. We can just add something like random_choice_dimension='DB' when we create a DARCDesign and it will randomly pick a fixed value for DB and do design optimisation over all remaining designs.