matthew29tang / TransporTA

CS170 Final Project - A greedy approximation solver for NP-hard variant of Travelling Salesman Problem

CS 170 Fall 2019 Project

Matthew Tang, Christianna Xu: Group 117

Usage

• Generate outputs: py .\solver.py --all inputs outputs (~5min w/ multicore)
• Local Autograder: py .\autograder.py --all inputs outputs (~7min w/ multicore, caching)
• Superscorer: py .\superscorer.py inputs outputs_folder1 ... outputs_foldern (~1.5hrs)
• Add --disableMulticore to disable Multicore for autograder and solver
  • Note that Multicore supresses runtime errors
• Add -v for verbose autograder output

Scores

Footsteps v0: 40.7345 (Deprecated)

Footsteps v2: 39.28

Footsteps v3: 39.273

Footsteps v3.5: 43.645

Footsteps v4: 39.019

Christofides v1: 60.780

Final superscore: 38.74 (Top 22% on leaderboard)

Instructions:

• Footsteps is the default solver. To switch to Christofides, uncomment out and return the solve call in solver.py
• Footsteps v4 is the final version
• Footsteps v2 uses smart_output instead of smarter_output. See the final report for more details.
• Footsteps v3 does not use a saturated set. See the final report for more details.
• Footsteps v3.5 had a bug where the newHomes was incorrectly calculated. But somehow, some of the outputs were better so the outputs were still superscored. See the final report for more details.