Refactor: add transforms and transformed emulators

[sgreenbury]

trafficstars

Closes #348.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 91.40271% with 38 lines in your changes missing coverage. Please review.

Project coverage is 78.11%. Comparing base (5e90347) to head (ef26951).

Files with missing lines	Patch %	Lines
...emulate/experimental/emulators/transformed/base.py	85.39%	13 Missing :warning:
autoemulate/experimental/transforms/utils.py	67.85%	9 Missing :warning:
autoemulate/experimental/transforms/vae.py	91.42%	6 Missing :warning:
autoemulate/experimental/transforms/base.py	90.69%	4 Missing :warning:
autoemulate/experimental/transforms/pca.py	90.62%	3 Missing :warning:
autoemulate/experimental/transforms/standardize.py	90.00%	3 Missing :warning:

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #474      +/-   ##
==========================================
+ Coverage   77.51%   78.11%   +0.59%     
==========================================
  Files         121      130       +9     
  Lines        8758     9198     +440     
==========================================
+ Hits         6789     7185     +396     
- Misses       1969     2013      +44     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

:rocket: New features to boost your workflow:

• :snowflake: Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions[bot]]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
autoemulate/experimental/emulators
__init__.py
base.py
autoemulate/experimental/emulators/neural_processes
conditional_neural_process.py
autoemulate/experimental/emulators/transformed
base.py					180-182, 351-352, 375-378, 383-388, 392-396
autoemulate/experimental/transforms
__init__.py
base.py					35-36, 43-44
pca.py					53-55
standardize.py					45-47
utils.py					60, 67-75
vae.py					90, 118-122, 142-144
tests
test_compare.py
tests/experimental
conftest.py
test_experimental_conditional_neural_process.py
test_experimental_transformed.py
tests/experimental/transforms
test_transforms.py
Project Total

This report was generated by python-coverage-comment-action

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[sgreenbury]

Following discussion:

• It would be good to double check if implications of using torch Transform for a VAE since the call and inv methods do not guarantee $f^{-1}(f(x)) = x$
• make_positive_definite might be useful to pass params too but we can revisit if there are issues that arise in using
• Subclassing ComposeTransform to AutoEmulateComposeTransform might benefit the API but this could potentially be considered as part of #531 if it seems useful in the top-level API?

[radka-j]

One last minor thought I had was about whether we want the input args to the inverse transform methods to be z rather than x or y? It might just be easier to track what's what but I am also more than happy to be told that's wrong.

[sgreenbury]

One last minor thought I had was about whether we want the input args to the inverse transform methods to be z rather than x or y? It might just be easier to track what's what but I am also more than happy to be told that's wrong.

Yeah I think this could definitely be clearer, perhaps the following refactor would be useful:

• Since torch's Transform refers to x in the domain and y in the codomain, I think the AutoEmulateTransform might be best if consistent with this and update the input API to x for call and y for inverses
• For the TransformedEmulator, since this works with x for emulator inputs and y for targets, using z to refer to inputs in the codomain of the transforms for x seems like a good option.
• Some of the method names in TransformedEmulator might be worth revisiting too (e.g. some refer to y like _inv_transform_y_mvn while the y transforms in the attribute are called target_transforms) - perhaps we could rename the y in method names to target for consistency. Or alternativelu target_transforms to y_transforms.

[sgreenbury]

@radka-j: thanks for the review comments! I've aimed to address these now and also more generally:

• Added docstrings for all _methods in TransformedEmulator and AutoEmulateTransform
• Renamed function arguments / attributes in AutoEmulateTransform and TransformedEmulator
• Renamed methods in AutoEmulateTransform for clarity

[sgreenbury]

Thanks @radka-j!