The Dantzig-Fulkerson-Johnson TSP formulation is easy to solve for few subtour constraints

This is GitHub repository accompaining the paper "The Dantzig-Fulkerson-Johnson TSP formulation is easy to solve for few subtour constraints".

The folder contains the following files:

• tsplib_hardtsplib.ipynb for reproducing the results of Table 1, 3 and 4 when you have small instances.
• tsplib_hardtsplib.py for reproducing the results of Table 1, 3 and 4 when you have large instances.
• focus_dantzig42_brazil58.ipynb for the results of Table 2 and Figure 3.
• random.ipynb for the results of Figure 4 and 5.
• plot.ipynb for the results of Figure 6.
• rectilinear_3D_instances.ipynb for the results of Section 5.4.
• utils.py, ialg.py, cover.py containing the functions used in the notebooks, that can be of independent interest.

To run the notebooks, first put in the folder data and put there the instances of the TSPLIB [1] and the HardTSPLIB [2]. Output file will be saved in the folder output. Here, you find the results of the experiments we have run.

[1] Reinelt, Gerhard. "TSPLIB—A traveling salesman problem library." ORSA journal on computing 3.4 (1991): 376-384.

[2] Vercesi, Eleonora, et al. "On the generation of metric TSP instances with a large integrality gap by branch-and-cut." Mathematical Programming Computation 15.2 (2023): 389-416.