Simplified neural network architecture for shortest path planning in optical network

The shortest path planning in optical networks is an integer linear programming problem (ILP). Since, this problem is NP hard therefore; a simplified scheme is required for real time solution. To counteract the problem, a simplified artificial neural network architecture is proposed which utilizes the information of the optical network for the initialization of neural networks weight parameter. In addition to this, a decision block has been incorporated in architecture which enhances the learning and quickly leads towards the steady state solution. The proposed strategy is tested on a 5 node network and results are found to be satisfactory.