Recent Advances in Safety Assurance of Neural Networks