baedan

i feel like we're talking past each other, and there's gotta be some concepts that we are defining very differently, haha

thanks for the responses. i'm tapped out for the day, will think about it more tomorrow

i think i understand the disconnect now. a policy in this package is not just the classically defined, stationary map from a state to a probability distribution over the action...

well, i can't wait for the next release. :D one thing though: what i mean by _policy evaluation_ or _prediction_ is not `(p::AbstractPolicy)(env)`, but evaluating the state/action values of a...

thanks, i'll look into it. all in all, this thread did make me consider more deeply the various aspects of design, which are indeed challenging.

> In fact, we also need contributors to work on porting [tablar methods](https://github.com/JuliaReinforcementLearning/ReinforcementLearning.jl/tree/v0.10.1/src/ReinforcementLearningZoo/src/algorithms/tabular) in the latest workflow in the master branch. would be great if there’s a document i can...

+1 to this. the error is hard to make sense of also

thank you for your help! > The most reliable option tends to be ForwardDiff over Zygote (as e.g. in `Zygote.hessian`). Some people also try mixing ReverseDiff & Zygote. will take...

haven't checked correctness yet, but happy to report that a second-order `Forward.gradient` on a first-order Zygote `gradient`, using `restructure`. using Zygote for both yields error `Type NamedTuple has no field...

i see, thanks for the explanation. i'm considering making a trigger resolution system that process entities representing triggered abilities. still brainstorming idea, so this is pretty rough. i've come up...