Add MultiViewLightGBM model

[tereshchuk1]

Description

Add MultiViewLightGBM model. Supports gene expression, methylation, mutations, copy number variation, and proteomics and drug fingerprints

Changes

• This comment contains a description of changes (with reason)
• If you've fixed a bug or added code that should be tested, add tests!

New features

• Added MultiViewLightGBM model
• Registered model in drevalpy/models/__init__.py
• Added hyperparameters to hyperparameters.yaml
• Added lightgbm as optional dependency in pyproject.toml
• Added lightgbm to nox test session extras in noxfile.py
• Added MultiViewLightGBM to baseline tests

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 87.50000% with 13 lines in your changes missing coverage. Please review.
✅ Project coverage is 80.92%. Comparing base (7d24cd6) to head (89f0cab).
⚠️ Report is 13 commits behind head on development.

Files with missing lines	Patch %	Lines
drevalpy/models/baselines/multi_view_lightgbm.py	87.00%	13 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@               Coverage Diff               @@
##           development     #430      +/-   ##
===============================================
+ Coverage        80.34%   80.92%   +0.58%     
===============================================
  Files              101      103       +2     
  Lines             8171     8336     +165     
===============================================
+ Hits              6565     6746     +181     
+ Misses            1606     1590      -16     

☔ View full report in Codecov by Harness.
📢 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.

[JudithBernett]

Comment by Claude because I'm not an expert on LightGBM and its standard parameters, but maybe worth looking into:

num_leaves vs max_depth are in conflict
This is the main issue. LightGBM is a leaf-wise tree grower, so num_leaves is the primary complexity control — max_depth is secondary and mainly used as a guardrail. The rule of thumb is:
num_leaves < 2^max_depth
With max_depth: 10, you could have up to 2^10 = 1024 leaves. Your num_leaves values of 63 and 127 are well within that, so there's no hard conflict — but if you intend max_depth: 10 to constrain complexity, it's effectively doing nothing here since num_leaves is already much smaller. You'd be better off either:

• Dropping max_depth and just tuning num_leaves, or
• Setting max_depth to something tighter like 6 or 7 if you actually want it to bite

num_leaves: 127 can be aggressive
127 leaves is fairly complex — fine for large datasets, but if your dataset is small/medium you may be giving the model too much capacity. Worth adding a smaller value like 31 to the search.

reg_lambda: 0.1 only
You're searching reg_alpha over two values but reg_lambda (L2) over only one. LightGBM defaults reg_lambda to 0.0, so 0.1 already adds regularization — but it's worth also trying 0 and maybe 1 here for symmetry with how you're treating reg_alpha.

[JudithBernett]

Just seeing this now: any particular reason why the xgboost, lightgbm, and gseapy requirements HAVE to be this version? I'd prefer >= for easier maintenance.

[PascalIversen]

Thanks! Could you also run the model and add it to the leaderboard?