This repository contains the code for the load balancing solutions of the virtual network. The main algorithm is Deep Q Network. ILP and Heuristic are used to compare experimental performance.

The Integer Linear Programming is implemented mainly in the ILP directory. The simple_steiner_tree.py is the solution for shortest path and the load_balancing_steiner_tree.py is the solution for load balancing. The Graph_generator.py is used to generate the network model.

The Heuristic is mainly used with k-shortest path (k = 1, 2, 3 ...). Choose the optimal path based on different metrics. Heuristic is implemented mainly in the Heuristic directory.

The main work is implemented using Deep Q Network to balance the resource of the network. In the DQN directory, the Grid2D_dqn_run.py is used to train the DQN model. The DQN model's network layer is in the Network directory.