Refactor sklearn models for experimental

[edwardchalstrey1]

• Based off Add refactored LGBM model to experimental emulators #399 - merge this first
• Closes Refactor: Create SklearnBackend class inheriting from Emulator base class & add first example (svm) #405

Questions:

• What are the full set of sklearn models that we want to keep?
  • Just SVM and Random Forest for now
• Do we actually need the SklearnBackend?
• If we do keep SklearnBackend, should LightGBM use this?
  • Probably not, if it is sufficiently similar then maybe - have a look at the docs. If we did, should call SklearnBackend something different
• Why are the checks in each model fit slightly different and do they need to be? If they can be standardised, we can move them into the fit method of SklearnBackend - if not, perhaps keep them in the specific model's fit func
• Add type hints to model args

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
autoemulate/emulators
gaussian_process.py
autoemulate/experimental/data
utils.py					176-178, 184
autoemulate/experimental/emulators
__init__.py
base.py					77-81, 193-195, 208-209, 217
lightgbm.py
random_forest.py
svm.py
autoemulate/experimental/emulators/gaussian_process
exact.py
tests
test_compare.py
tests/experimental
test_experimental_base.py
test_experimental_compare.py
test_experimental_random_forest.py
test_experimental_svm.py
Project Total

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

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 93.10345% with 12 lines in your changes missing coverage. Please review.

Project coverage is 80.37%. Comparing base (25d69e7) to head (93c3856).

Files with missing lines	Patch %	Lines
autoemulate/experimental/emulators/base.py	80.48%	8 Missing ⚠️
autoemulate/experimental/data/utils.py	77.77%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #415      +/-   ##
==========================================
+ Coverage   80.09%   80.37%   +0.28%     
==========================================
  Files         100      104       +4     
  Lines        6902     7057     +155     
==========================================
+ Hits         5528     5672     +144     
- Misses       1374     1385      +11     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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

[sgreenbury]

Adding the code from discussion with @edwardchalstrey1 and @radka-j on possible revision to the base class (relating to #322):

class Emulator(ABC, InputTypeMixin):
    """
    The interface containing methods on emulators that are
    expected by downstream dependents. This includes:
    - `AutoEmulate`
    """

    @abstractmethod
    def __init__(
        self, x: InputLike | None = None, y: InputLike | None = None, **kwargs
    ): ...

    @classmethod
    def model_name(cls) -> str:
        return cls.__name__

    @abstractmethod
    def _fit(self, x: InputLike, y: InputLike | None): ...


    @abstractmethod
    def check(self, x: InputLike, y: InputLike | None): ...

    @abstractmethod
    def convert(self, x: InputLike, y: InputLike | None) -> tuple[InputLike, InputLike | None]: ...

    def fit(self, x, y):
        """
        Fit the model to the data.

        Parameters
        ----------
            x: InputLike
                Input features as numpy array, PyTorch tensor, or DataLoader.
            y: OutputLike or None
                Target values (not needed if x is a DataLoader).

        Returns
        -------
            None
        """
        x, y = self.convert(x, y)
        self.check(x, y)
        self._fit(x, y)

    @abstractmethod
    def predict(self, x: InputLike) -> OutputLike: ...

    @staticmethod
    @abstractmethod
    def get_tune_config() -> TuneConfig: ...

[sgreenbury]

Added this here as initial functionality form preprocessing outputs. This would be good to revisit in #348 where this might be better as distinct functionality that can be combined with the model and/or updated with the API used there.

[sgreenbury]

Also relates to #437

[edwardchalstrey1]

Note @sgreenbury I have added tests for these in #464

[sgreenbury]

As discussed, this looks good to merge with subsequent changes with checks/validation to follow in #422, thanks @edwardchalstrey1!