Simultaneous Multi-Category Binning Optimization

[akhter-towsifa]

Currently, the binning optimization algorithm handles multi category cases in the following way:

• takes the most sensistive category (i.e. res2b) in a given channel and finds the optimal binning with low expected limit,
• freezes this optimal binning and uses it as a starting point for the next category (i.e. res1b),
• calculates the best binning for the current category (i.e. res1b), freezes this one and uses it as a starting point for next category (i.e. boosted, and so on).

The current algorithm assumes some correlation in kinematics between the categories, but we want a more rigorous option that will optimize all the categories in a channel simultaneously.

• This approach will be computationally heavy (evaluating limits over multiple categories at a time)
• But will create a more consistent binning scheme throughout different categories.

[Copilot]

Pull request overview

This pull request implements simultaneous multi-category binning optimization for statistical inference, moving from a sequential approach where categories are optimized one-by-one to a simultaneous approach where all categories in a channel are optimized together. This is computationally heavier but produces more consistent binning schemes across categories.

Changes:

• Adds new multi_optimize_binning.py module with MultiBinning and MultiBayesianOptimization classes for simultaneous multi-category optimization
• Updates optimize_channel.py to support simultaneous multi-category optimization mode controlled by configuration
• Replaces deprecated distutils.util.strtobool with a custom implementation in optimize_binning.py
• Adds multi_category_optimization flag to bin_optimization.yaml configuration

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 16 comments.

File	Description
bin_opt/multi_optimize_binning.py	New file implementing simultaneous multi-category binning optimization with MultiBinning class and MultiBayesianOptimization algorithm
bin_opt/optimize_channel.py	Adds multi-category optimization path, new helper functions for multi-category result selection and input file discovery
bin_opt/optimize_binning.py	Replaces deprecated distutils.util.strtobool with custom implementation for Python 3.12+ compatibility
bin_opt/bin_optimization.yaml	Adds multi_category_optimization configuration flag (set to True)

Comments suppressed due to low confidence (1)

bin_opt/optimize_channel.py:286

• The variable other_cat_file is only assigned inside the inner loop (line 284) if a matching file is found. If no matching file exists in best_dir, the variable will be undefined when checked at line 286, causing a NameError. The check should be inside the loop or a default value should be set before the loop.

        for cat_idx in range(cat_index):
            cat = categories[cat_idx][0]
            for file in os.listdir(best_dir):
                if file.endswith('.txt') and cat in file:
                    other_cat_file = f"{best_dir}/{file}"

            if not os.path.isfile(other_cat_file):

---

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