/DeepQAP

Deep RL agent that solves Quadratic Assignment Problems

Primary LanguageJupyter Notebook

Deep QAP

This repo contains the code of a Deep Reinforcement Learning agent that is trained to solve Quadratic Assignment Problems (QAP).

For details and results see the thesis paper. As an introduction to the problem and setup, see the proposal presentation.

Setup

Requires python 3.9

Install poetry

pip install poetry

Setup environment

poetry install
poetry shell

Training runs

From within the environment execute either one of the run_* scripts or one of the scripts in scripts/*.

Jupyter evaluation notebooks

Plots and small experiments.

Within the environment, run

jupyter notebook

then open the browser interface and open a notebook in the evaluation folder

Bokeh evaluation server

Shows some information about the trained agent and allows interactive execution of the agents.

bokeh serve evaluation/server

Run unit tests

Only cover some critical aspects

pytest