1: \begin{abstract}

2:  We establish lower bounds on the complexity of finding

3:  $\epsilon$-stationary points of smooth, non-convex

4:  high-dimensional functions using first-order methods. We prove 

5:  that deterministic first-order methods, even applied to arbitrarily smooth

6:  functions, cannot achieve convergence rates in $\epsilon$ better than

7:  $\epsilon^{-8/5}$, which is within $\epsilon^{-1/15}\log\frac{1}{\epsilon}$ of 

8:  the best known rate for such methods. Moreover, for functions with Lipschitz 

9:  first and second

10:  derivatives, we prove no deterministic first-order method can achieve 

11:  convergence rates better than

12:  $\epsilon^{-12/7}$, while $\epsilon^{-2}$ is a lower bound for

13:  functions with only Lipschitz gradient. For \emph{convex} functions with 

14:  Lipschitz gradient, accelerated gradient descent achieves the rate 

15:  $\epsilon^{-1}\log\frac{1}{\epsilon}$, showing that finding stationary points 

16:  is  easier given convexity.

17: \end{abstract}

18: