ebrahimpichka / QAP-RL

Unofficial implemnetation of "Solving Quadratic Assignemt Problem using Deep Reinforcement Learning" (https://arxiv.org/abs/2310.01604)

Solving Quadratic Assignemt Problem using Deep Reinforcement Learning

This reposiroty contains an unofficial implemnetation of the model proposed in Bagga, P. S., & Delarue, A. @adelarue (2023). Solving the quadratic assignment problem using deep reinforcement learning. for solving QAP via Deep RL.

The paper formulizes QAP as a sequential decision problem (i.e. an MDP) by selecting locations and their assigned facilitiy sequentially and proposes a Double Pointer Network architecrure to represent the actor to learn to solve the problem using the in the Advantage Actor-Critic algorithm .

Reference

[1] Bagga, P. S., & Delarue, A. (2023). Solving the quadratic assignment problem using deep reinforcement learning. https://doi.org/10.48550/ARXIV.2310.01604

TODO: add action masking to loc/fac selection TODO: complete A2C training loop