tsp-200

Metric tour-length, minimize · Best known none recorded · Play · Code · All example problems

hillclimb problem get tsp-200      # copies the problem into hillclimb/problems/
hillclimb verify tsp-200           # scores the floor; spread 0, the verifier is exact
hillclimb run tsp-200 --budget 15m

The problem

data/cities.csv contains 200 cities (id,x,y, coordinates in the unit square, fixed instance). Find a closed tour visiting every city exactly once with the shortest possible total Euclidean length (the tour returns from the last city to the first).

This instance is fixed and seeded — the score is fully reproducible. A random order is around 100; nearest-neighbor lands near 12; good local search (2-opt / Or-opt, candidate lists, restarts or simulated annealing) pushes toward ~10. numpy, scipy and pandas are available.

Submission format

Write submission.csv in the working directory with the header position,city: 200 rows where position = 0..199 and city is a permutation of the city ids 0..199, in visiting order. sample_submission.csv is the identity order (a weak valid baseline).

Scoring

The orchestrator runs problem/verify.py after your script finishes. The verifier validates the tour and prints val_score: <tour length> (or a large penalty score of 1000.0 if the tour is invalid). Lower is better.

There is no train/test split; this is a pure optimization problem. Keep total runtime well within the execution time limit.