#Python Plotter for Constraint Optimization Results

This directory contains tools for plotting results from constraint optimization experiments.

#Files

• plotter.py - Main plotting script
• config_example.json - Example configuration file
• config_feasible_fraction.json - Example config for feasible fraction plots
• config_constraint_violation.json - Example config for constraint violation plots
• objectives.json - Global configuration with optimal objective values

#Configuration

The JSON configuration specifies:

• plot_type: Type of data to plot (best_candidates, constraint_violation, feasible_fraction)
• algorithms: List of algorithms with visualization settings
  • name: Algorithm directory name
  • label: Display name in legend
  • color: Algorithm color (hex code)
  • linestyle: Line style (-, --, -., :)
• instances: List of problem instances with colors
  • name: Instance directory name
  • label: Display name in legend
  • color: Instance color (hex code)
• data_path: Path to solutions directory
• output_dir: Directory for saving plots
• plot_settings: Figure settings (size, labels, grid, log scales, etc.)
  • log_x: Enable logarithmic x-axis
  • log_y: Enable logarithmic y-axis
  • show_std: Show standard deviation bands (default: true)
  • color_combination: How to combine algorithm and instance colors
    • "blend": Blend algorithm and instance colors (default)
    • "algorithm": Use only algorithm colors
    • "instance": Use only instance colors

#Usage

Basic usage:

python3 plotter.py config_example.json

With Guix (recommended):

guix shell python python-matplotlib python-pandas python-numpy -- python3 plotter.py config_example.json

#Data Structure

The plotter expects data in this structure:

data_path/
├── algorithm1/
│   ├── instance1/
│   │   ├── best_candidates_*.csv
│   │   ├── constraint_violation/
│   │   │   └── constraint_violations_*.csv
│   │   └── feasible_fraction/
│   │       └── feasible_fractions_*.csv
│   └── instance2/...
└── algorithm2/...

#Features

• Averages multiple runs automatically
• Shows standard deviation as transparent bands around mean
• Single plot with multiple instances (colored by instance)
• Different line styles for algorithms
• Logarithmic axes support for both x and y
• Percentage deviation from optimal for best_candidates plots
• Absolute values for other plot types (constraint_violation, feasible_fraction)
• Automatic legend with run counts
• High-quality SVG output

#Percentage Deviation Calculation

For best_candidates plots, the plotter automatically:

1. Loads optimal objective values from objectives.json
2. Validates that all values are >= optimal (raises error if better values found)
3. Calculates percentage deviation: (current - optimal) / |optimal| * 100
4. Plots deviation percentages so you can see convergence to 0%

Other plot types use absolute values as normal.