How adrianfaz/fair-representation-in-submodular-subset-selection-a-pareto-optimization-approach works
Project type and purpose
• This is an open-source research tool (CLI application) for fair submodular subset selection via Pareto optimization.
• The README title and filenames (run_mc_*, run_rs_*, data folders) show it’s geared toward running different submodular algorithms on various datasets and producing result files.
File structure
• Top-level “run_*.py” scripts: entry points for experiments (Monte Carlo (MC) and Randomized Swap (RS) methods on different constraints).
• algos/: core algorithm implementations that operate on data matrices.
• handlers/: classes (Handler.py, HandlerMC.py, HandlerRS.py, etc.) that orchestrate an experiment: loading data, invoking the right algorithm, collecting metrics.
• tools/: shared utilities (e.g. data preprocessing, logging).
• data/: raw and preprocessed datasets (dblp, facebook, wine, ml), in CSV/pickle form.
• results/: output directory where each run script dumps its CSV results.
README for additional insights
• The README is minimal; it confirms “All data and software.” We infer that the primary architecture emerges from the folder layout and script naming.
Elements
• From Runner → Handler
• Handler → Data Store (read)
• Handler → algos module
• algos → Handler (return results)
• Handler → Results Store (write)
• Handler ↔ Tools (utilities calls)
• Rectangles for code modules (handlers/, algos/, tools/)
• Rounded rectangles for scripts (run_*.py entry points)