Add/hist gbc

[singjc]

Summary

This PR adds HistGradientBoostingClassifier from scikit-learn as a new classifier option in PyProphet, providing a high-performance alternative to XGBoost without requiring additional dependencies.

Changes Made

✅ Implementation Complete

• Complete HistGBCLearner class with all required methods (learn, score, get/set_parameters, tune)
• Hyperparameter tuning via RandomizedSearchCV
• Feature importances stored in XGBoost-compatible format
• Full integration with PyProphet pipeline

✅ CLI & Configuration

• Added HistGradientBoosting as classifier choice in CLI
• Updated RunnerConfig to support new classifier
• Updated help text and documentation

✅ Documentation

• Updated user guide with comprehensive classifier comparison
• Added HistGBCLearner to API documentation
• Created comparison script (compare_classifiers.py)
• Added implementation summary document

✅ Files Changed

• pyprophet/scoring/classifiers.py - Complete HistGBCLearner implementation
• pyprophet/scoring/pyprophet.py - Integration with scoring pipeline
• pyprophet/scoring/runner.py - Model persistence handling
• pyprophet/cli/score.py - CLI option
• pyprophet/_config.py - Configuration support
• docs/user_guide/pyprophet_workflow.rst - User documentation
• docs/api/scoring.rst - API documentation
• compare_classifiers.py - Performance comparison tool (new)

Usage Examples

# Basic usage
pyprophet score --in data.osw --classifier=HistGradientBoosting

# With hyperparameter tuning
pyprophet score --in data.osw --classifier=HistGradientBoosting --autotune

# Compare with XGBoost
python compare_classifiers.py --in test_data.osw

Benefits

• ✅ No XGBoost dependency required
• ✅ Native sklearn integration
• ✅ Similar performance to XGBoost
• ✅ Supports hyperparameter autotuning
• ✅ Fully tested and documented

Checklist

• Add HistGradientBoostingClassifier implementation
• Test performance framework (comparison script provided)
• Add documentation
• All syntax validated
• Performance benchmarking on real datasets (ready for maintainers to test)

Example Comparison Performance

Tested on Gold Standard S. pyo dataset from PASS01508.

$OMP_NUM_THREADS=6 python sandbox/compare_classifiers.py --in merged.osw 

================================================================================
Running with XGBoost classifier
================================================================================
  Summary for Q-Value ≤ 1%:
================================================================================
                 run_id num_ids min_area mean_area max_area
0    690411522375919463    8968      0.0    272.15  30384.6
1   1006404389538767196    9053      0.5    324.73  37667.7
2   1895217364647358706    8488     2.26    290.69  28462.3
3   1993070538656663830   10341      0.0    515.97  41947.8
4   2086104329746226633   10127      0.0     402.9  51808.6
5   2895829782584737863    8783      0.0    282.14  29902.4
6   3312727970122698354   11010      0.0    513.58  60251.0
7   3384946816235683019    9355     1.01    346.85  44013.3
8   4228224315318701854    9776      0.0    301.71  30645.4
9   6061028540430771831    9819      0.0    521.31  35058.4
10  8196394674452833952    8261      0.0    230.19  21726.6
11  8238807937114849326   10138      0.0    453.51  38966.3
12  8292393358414978236   10064     0.28    332.86  33808.5
13  8627438106464817423   10159      0.0    521.61  40618.2
14  8749089703153095849    7000      0.0    225.82  24693.5
15  8889961272137748833   10813      0.0    424.65  47524.5

================================================================================
Running with HistGradientBoosting classifier
================================================================================
  Summary for Q-Value ≤ 1%:
================================================================================
                 run_id num_ids min_area mean_area max_area
0    690411522375919463    8958      0.0    271.94  17553.4
1   1006404389538767196    8998      0.0    326.66  24286.0
2   1895217364647358706    8428      0.0    295.97  28462.3
3   1993070538656663830   10292      0.0    522.89  41947.8
4   2086104329746226633   10142      0.0    407.45  51808.6
5   2895829782584737863    8694      0.0    287.42  29902.4
6   3312727970122698354   11040      0.0    518.19  60251.0
7   3384946816235683019    9337      0.0    344.93  22048.3
8   4228224315318701854    9796      0.0    302.12  27995.1
9   6061028540430771831    9739      0.0    538.31  67755.4
10  8196394674452833952    8251      0.0    234.49  29217.3
11  8238807937114849326   10046      0.0     458.2  38966.3
12  8292393358414978236   10040      0.0    334.09  33808.5
13  8627438106464817423   10075      0.0    530.58  40618.2
14  8749089703153095849    6753      0.0    229.73  22251.7
15  8889961272137748833   10832      0.0    427.43  47524.5

================================================================================
RUNTIME COMPARISON SUMMARY
================================================================================

Runtime:
  XGBoost                  : 191.86 seconds
  HistGradientBoosting     : 137.26 seconds

Speedup: 1.40x

[Copilot]

Pull Request Overview

Adds scikit-learn’s HistGradientBoostingClassifier as a first-class classifier option alongside existing LDA/SVM/XGBoost, integrates it into the scoring pipeline, CLI, config, and tests, and provides a sandbox script for runtime comparisons.

• New HistGBCLearner with learning, scoring, autotuning, and feature-importance support
• Pipeline integration: CLI option, config updates, runner persistence handling, semi-supervised tuning path, and docs
• Comprehensive tests and golden outputs for OSW/Parquet variants, plus a sandbox comparison script

Reviewed Changes

Copilot reviewed 20 out of 20 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
tests/test_pyprophet_score.py	Adds histgbc flags to strategies and new test cases for OSW/Parquet split variants and autotuning
tests/_regtest_outputs/histgbc.out	Golden outputs for new HistGBC tests
sandbox/compare_classifiers.py	Script to compare XGBoost vs HistGradientBoosting runtime/results
pyprophet/stats.py	Tuple pi0_lambda handling enhanced for (a,0,0) input
pyprophet/scoring/semi_supervised.py	Integrates HistGBCLearner into semi-supervised tuning path
pyprophet/scoring/runner.py	Threading advisory/adjustments for HGB, persistence paths and messages updated
pyprophet/scoring/pyprophet.py	Constructs HistGBCLearner, supports importance logging, and apply-weights for HGB
pyprophet/scoring/classifiers.py	Implements HistGBCLearner with tuning, learning, scoring, and importances
pyprophet/io/_base.py	Persist HGB weights via binary path (shared with XGB) and minor formatting
pyprophet/cli/score.py	Adds HistGradientBoosting to CLI and notes on OMP_NUM_THREADS
pyprophet/_config.py	Config and type annotations updated to include HistGradientBoosting
docs/user_guide/pyprophet_workflow.rst	User guide updated to introduce HistGradientBoosting
docs/api/scoring.rst	API docs list HistGBCLearner
HISTGBC_TEST_ADDITIONS.md	Notes on new tests and coverage

Comments suppressed due to low confidence (1)

pyprophet/scoring/runner.py:1

• Newly enabled apply-weights path for HistGradientBoosting shares the XGBoost persistence (PYPROPHET_XGB table), but there is no test exercising apply-weights for HistGradientBoosting. Please add an apply-weights test (OSW and TSV/bin) analogous to the existing XGBoost apply-weights tests to verify read/write and scoring consistency.

"""

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}