This repository contains the work for the Multi-Source Prim-Dijkstra contest "entry". In the contest folder:
-
mspd.py: Our implementation of the MSPD heuristic in accordance with the original implementation in C++ here. This is primarily used to produce objective values (i.e., normalized wirelength and skew) in the agent procedures.
-
local_search.py: Our implementation of local search.
-
qlearning.py: Our implementation of Q-Learning with local search refinement.
-
inference.py: Provided driver code for the contest.
-
contest.md: Provided description of the contest.