Skip to content

Conversation

penelopeysm
Copy link
Member

@penelopeysm penelopeysm commented Jan 24, 2025

This PR adds a series of new docs on:

  • why variables might need to be transformed;
  • how this is accounted for mathematically with the Jacobian term;
  • basic API and functionality of Bijectors.jl;
  • higher-level usage of transforms in DynamicPPL and Turing.

Comments are welcome:)

PS. I know that there is some subtlety with the Turing.jl HMC sampling section because the print statements don't actually happen during the sampling, they occur when values_as_in_model is called (post-sampling) to collect the values. I figured it's ok to gloss over that, though, because it's still true that HMC sampling occurs entirely in unconstrained space.

Copy link
Contributor

Preview the changes: https://turinglang.org/docs/pr-previews/582
Please avoid using the search feature and navigation bar in PR previews!

Copy link
Contributor

@willtebbutt willtebbutt left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just minor stuff. I like these docs a lot!

Copy link
Member

@mhauru mhauru left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is great! I had a few small comments, but they are minor enough that I'm happy to merge as-is if need be.

@penelopeysm
Copy link
Member Author

I'll leave this up a few more days to see if anybody has more input before merging:)

@penelopeysm penelopeysm merged commit 2a44076 into master Feb 3, 2025
2 checks passed
@penelopeysm penelopeysm deleted the py/transforms branch February 3, 2025 15:21
github-actions bot added a commit that referenced this pull request Feb 3, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants