Add learn-weights CLI step with PU learning and Optuna tuning

[iankchristie]

Introduces a new reVeal learn-weights pipeline step that trains a PUExtraTrees model on a normalized grid using point labels as positive samples. Derives feature importance weights for use with score-weighted.

• Add reVeal/pu/ — PUExtraTree/PUExtraTrees adapted from https://github.com/jonathanwilton/PUExtraTrees, with modifications from /projects/largeload/.../models/PUExtraTrees/ (deterministic seeding, joblib parallelism)
• Add reVeal/learn_weights.py — data preparation logic adapted from /projects/largeload/.../models/prepare_data.py (DataHandler), training and metrics from /projects/largeload/.../models/models.py (ModelTrainer)
• Add Optuna-based hyperparameter tuning for class_prior — adapted from ModelTrainer.tune_hyperparameters() and ModelTrainer.objective() in /projects/largeload/.../models/models.py
• Add CLI command (reVeal/cli/learn_weights.py), config (reVeal/config/learn_weights.py), and tests — new code following existing reVeal CLI patterns (score_weighted, normalize)
• Add joblib, scipy, optuna to pyproject.toml and environment.yml

Tested by running the new CLI tool like so

python -m reVeal.cli.cli learn-weights -c /projects/largeload/geospatial/runs/test_scenarios/learned_weights_2026-05-19_agg64/config_learn_weights.json

You can use the learned_weights_2026-05-19_agg64/ directory to test this code.

Next Steps:
This works well to generate the score weights. However, in practice there many of these features are highly correlated or we want to manually exclude for other purposes. We want a few feature engineering tools that will allow users to iterate on the score outputs. This will include:

1. An exclude list
2. Visualization features (dendrogram and correlation matrix) to help users feature engineer before it goes into the score_weighted step.

[Copilot]

Pull request overview

Adds a new reVeal learn-weights pipeline step that uses PU (positive–unlabeled) ExtraTrees training to derive feature-importance-based weights and emit a score-weighted-compatible configuration.

Changes:

• Introduces a PU ExtraTrees/ExtraTree implementation (reVeal/pu/*) with optional joblib parallelism and deterministic seeding.
• Adds core learning + Optuna tuning workflow (reVeal/learn_weights.py) and exposes it via a new CLI command/config (reVeal/cli/learn_weights.py, reVeal/config/learn_weights.py).
• Adds tests for data prep, training, tuning, and config generation (tests/test_learn_weights.py) and updates dependencies (pyproject.toml, environment.yml).

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 20 comments.

Show a summary per file

File	Description
tests/test_learn_weights.py	New test coverage for the learn-weights pipeline helpers and Optuna tuning.
reVeal/pu/trees.py	Adds PUExtraTrees forest implementation plus grid/cutpoint utilities.
reVeal/pu/tree.py	Adds PUExtraTree decision tree implementation used by the forest.
reVeal/pu/init.py	Exposes PUExtraTrees from the new reVeal.pu package.
reVeal/learn_weights.py	Implements PU data preparation, training, Optuna tuning, and weight/config generation.
reVeal/config/learn_weights.py	Adds pydantic config model for learn-weights inputs.
reVeal/cli/learn_weights.py	New CLI command wiring to run learn-weights and write outputs.
reVeal/cli/cli.py	Registers the new learn-weights command in the main CLI.
pyproject.toml	Adds new runtime dependencies needed by learn-weights/PU trees (but missing scikit-learn).
environment.yml	Adds conda dependencies for the new functionality (but missing scikit-learn).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 64.18605% with 77 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.62%. Comparing base (dc210e6) to head (2374e7b).

Files with missing lines	Patch %	Lines
reVeal/learn_weights.py	72.72%	35 Missing and 4 partials ⚠️
reVeal/cli/learn_weights.py	26.92%	38 Missing ⚠️

❌ Your patch status has failed because the patch coverage (64.18%) is below the target coverage (80.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #55      +/-   ##
==========================================
- Coverage   84.50%   81.62%   -2.89%     
==========================================
  Files          17       20       +3     
  Lines        1304     1518     +214     
  Branches      180      200      +20     
==========================================
+ Hits         1102     1239     +137     
- Misses        163      236      +73     
- Partials       39       43       +4     

Flag	Coverage Δ
unittests	81.62% <64.18%> (-2.89%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.

[ppinchuk]

Thanks for adding tests and keeping the environment lockfile updated!