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
As the title says, the depth parameter is not used by DPWSolver.
I dug a bit and found out that the function estimate_value defined in domain_knowledge.jl line 65 will use the parameter max_depth set by default to 50 instead of depth that one should set when calling DPWSolver. It will only use depth if max_depth = -1.
According to dpw_types.jl, estimate_value is set by default to RolloutEstimator(RandomSolver(rng)). Calling DPWSolver(estimate_value = RolloutEstimator(RandomSolver(rng), max_depth=-1)) would do the trick but it is not obvious in the doc that depth is useless if max_depth is not set to -1.
My issue here is not really an issue but did I miss something ? Should it be easier to use my depth parameter ? Or should it be in the doc ?
Thanks !
The text was updated successfully, but these errors were encountered:
Wow! Thanks for catching this very subtle issue. I think the behavior is correct (See #88 for more context), but the documentation is wrong. I will update it.
Hi !
As the title says, the depth parameter is not used by DPWSolver.
I dug a bit and found out that the function estimate_value defined in domain_knowledge.jl line 65 will use the parameter max_depth set by default to 50 instead of depth that one should set when calling DPWSolver. It will only use depth if max_depth = -1.
According to dpw_types.jl, estimate_value is set by default to RolloutEstimator(RandomSolver(rng)). Calling DPWSolver(estimate_value = RolloutEstimator(RandomSolver(rng), max_depth=-1)) would do the trick but it is not obvious in the doc that depth is useless if max_depth is not set to -1.
My issue here is not really an issue but did I miss something ? Should it be easier to use my depth parameter ? Or should it be in the doc ?
Thanks !
The text was updated successfully, but these errors were encountered: