Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Better names for sampling methods from trees #99

Closed
xukai92 opened this issue Sep 13, 2019 · 8 comments
Closed

Better names for sampling methods from trees #99

xukai92 opened this issue Sep 13, 2019 · 8 comments
Milestone

Comments

@xukai92
Copy link
Member

xukai92 commented Sep 13, 2019

See 26aabb5#commitcomment-34976303

@xukai92
Copy link
Member Author

xukai92 commented Sep 14, 2019

Related discussion: #91 (comment)

@yebai
Copy link
Member

yebai commented Sep 18, 2019

Maybe we can remove Slice and Multinomial from exports? These exports are only useful for expert-users, so we might not want to export them anyway.

@xukai92
Copy link
Member Author

xukai92 commented Sep 19, 2019

I prefer to rename them for several reasons.

  • Users have to use AdvancedHMC.Slice if we don't export them.
  • AdvancedHMC IS for "expert-users".

Some options:

  1. Go for the simplicity: Multinomial -> Multi
  2. Go for self-contained: TrajectoryMultinomialSampler and TrajectorySliceSampler

Any thoughts?

@xukai92
Copy link
Member Author

xukai92 commented Sep 19, 2019

Option 3:

Define @enum TrajectorySampler slice multinomial and somehow dispatch over them. (Might be tricky?)

@yebai
Copy link
Member

yebai commented Sep 19, 2019

I mean in most cases we will use the default NUTS configuration, so maybe it’s ok to remove Slice/Multinomial from exports. Also, the user can still explicitly import these names if needed. Then we can consider creating a sugar syntax, e.g. NUTS(...) will create a NUTS sampler with Multinomial sampler by default.

@yebai
Copy link
Member

yebai commented Sep 20, 2019

Alternatively, we can shorten SliceTrajectorySampler to SliceTS, similarly MultinomialTS and MetropolisTS. The meanings of these types should be clear from their common supertype AbstractTrajectorySampler.

@xukai92
Copy link
Member Author

xukai92 commented Sep 25, 2019

I like this ???TS proposal! Let's do this.

@xukai92 xukai92 added this to the aabi2020 milestone Oct 3, 2019
@yebai
Copy link
Member

yebai commented Oct 23, 2019

Closed by #114.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants