From paper figures to reproducible experiment outputs
A paper figure is a final artifact. A reproducible output is a traceable artifact.
When I reproduce a routing experiment, I need more than a figure that resembles the original paper. I need to know which input data, parameters, algorithm mode, decoded solution, and cost calculation produced that figure.
What I try to preserve:
- raw instance data
- cleaned data
- parameter settings
- algorithm configuration
- run index or seed
- runtime log
- best-so-far history
- feasibility messages
- route figure
- convergence curve
- result table
- failure notes
The goal is simple: when I look back at a figure or table, I should know which data, code, parameters, and run produced it.
A good figure is useful. But in a research system, the figure should be the surface of a traceable run, not an isolated screenshot.