this repository do the simulation of two classical methods to escape from the saddle point. The first one is Jin C, Ge R, Netrapalli P, et al. How to escape saddle points efficiently. The second one is Natasha 2: Faster non-convex optimization than sgd. These two simulation is accomplished using Python. We also do the time complexity test in the end. We test only for the quadratic function because of the convenience to calculate the gradient. We are pleased to find that both these methods contribute to escaping from the saddle point.