Towards End-To-End Learning-Based Algorithms In Motion Planning