Verification of neural networks: Enhancing scalability through pruning