How diran001/capstone-bbo works
This is a research workflow for sequential black-box Bayesian optimisation, not a general-purpose optimisation library. Each weekly cycle starts from cumulative observations, fits a function-specific Gaussian Process, selects a candidate using an acquisition rule, and prepares portal submissions; returned scores are registered for the next cycle. The repository also supports historical visualisation and Power BI analysis. The README documents the notebook-led workflow, while the sampled code shows the plotting engine and log-to-dataset export; notebook internals were not sampled, so those connections are documented rather than source-verified.
Main entry points: bbo_workflow.ipynb Weekly notebook · submissions Weekly candidates · processed Returned results · bbo_plotting_patch.py Enhanced plotting
Components 15
Optimisation workflow
- Weekly notebook
notebooks/bbo_workflow.ipynb - Weekly tuning
data/config/tuning_by_week.csv - GP surrogate
notebooks/bbo_workflow.ipynb - Acquisition selection
notebooks/bbo_workflow.ipynb
Experiment records
- Cumulative observations
data/initial_data - Weekly candidates
data/submissions - Returned results
data/processed - Execution logs
data/logs
Visualisation
- Historical plots
data/plots - Enhanced plotting
scripts/bbo_plotting_patch.py - Plot backfill
scripts/backfill_historical_enhanced_plots.py
Reporting
- Power BI export
scripts/build_powerbi_dataset.py - Power BI dataset
data/powerbi
Other components
- Researcher
- Course portal
Connections 17
- Researcher → Weekly notebook: runs weekly
README.md - Weekly notebook → Cumulative observations: reads observations
README.md - Weekly notebook → GP surrogate: fits surrogate
README.md - Weekly notebook → Weekly tuning: uses tuning
- Weekly notebook → Acquisition selection: selects candidate
README.md - Weekly notebook → Weekly candidates: writes inputs
README.md - Weekly candidates → Course portal: submits query
README.md - Course portal → Returned results: returns scores
README.md - Researcher → Returned results: places return
README.md - Returned results → Weekly candidates: registers outputs
README.md - Weekly notebook → Execution logs: records execution
README.md - Weekly notebook → Enhanced plotting: generates plots
scripts/bbo_plotting_patch.py - Enhanced plotting → Historical plots: saves figures
scripts/bbo_plotting_patch.py - Plot backfill → Enhanced plotting: calls plotting
scripts/backfill_historical_enhanced_plots.py - Plot backfill → Cumulative observations: reads arrays
scripts/backfill_historical_enhanced_plots.py - Power BI export → Execution logs: reads logs
scripts/build_powerbi_dataset.py - Power BI export → Power BI dataset: writes dataset
scripts/build_powerbi_dataset.py