Stabilizing the training of deep neural networks using Adam optimization and gradient clipping | IJSREM