LIME Interpretability

[naveenkcb]

Contributor: Naveen Baskaran

Contribution Type: Interpretability method, Tests, Example

Description
This PR implements the LIME (Local Interpretable Model-agnostic Explanations ) interpretability method for PyHealth models, enabling users to understand which features contribute most to model predictions. LIME explains model predictions by approximating the model locally with an interpretable surrogate model by creating feature perturbations and weighting locally.

Files to Review

1. pyhealth/interpret/methods/init.py - added the LIME explainer class
2. pyhealth/interpret/methods/lime.py - Core LIME method implementation. Supports embedding based attribution, continuous feature support
3. examples/lime_stagenet_mimic4.py - Example script showing the usage of LIME method
4. tests/core/test_lime.py - added comprehensive test cases to test the main class, utility methods and attribution methods.
5. docs/api/interpret.rst - added support for LIME documentation
6. docs/api/interpret/pyhealth.interpret.methods.lime.rst - added a new toctree file for LIME

[jhnwu3]

Looks like you accidentally duplicated the SHAP changes into the same PR. you'll need to resolve the merge conflict.

[naveenkcb]

@jhnwu3 - I fixed this PR's commits to keep only the LIME related changes. it is ready for review now.