## Abstract

We examine the behavior of accelerated gradient methods in smooth nonconvex unconstrained optimization, focusing in particular on their behavior near strict saddle points. Accelerated methods are iterative methods that typically step along a direction that is a linear combination of the previous step and the gradient of the function evaluated at a point at or near the current iterate. (The previous step encodes gradient information from earlier stages in the iterative process). We show by means of the stable manifold theorem that the heavy-ball method is unlikely to converge to strict saddle points, which are points at which the gradient of the objective is zero but the Hessian has at least one negative eigenvalue. We then examine the behavior of the heavy-ball method and other accelerated gradient methods in the vicinity of a strict saddle point of a nonconvex quadratic function, showing that both methods can diverge from this point more rapidly than the steepest-descent method.

This is a preview of subscription content, access via your institution.

## References

Attouch, H., Cabot, A.: Convergence rates of inertial forward–backward algorithms. SIAM J. Optim.

**28**(1), 849–874 (2018)Attouch, H., Goudou, X., Redont, P.: The heavy ball with friction method, I. The continuous dynamical system: global exploration of the local minima of a real-valued function by asymptotic analysis of a dissipative dynamical system. Commun. Contemp. Math.

**2**(01), 1–34 (2000)Attouch, H., Peypouquet, J.: The rate of convergence of Nesterov’s accelerated forward-backward method is actually faster than \(1/k^2\). SIAM J. Optim.

**26**(3), 1824–1834 (2016)Bubeck, S.: Convex optimization: algorithms and complexity. Found. Trends Mach. Learn.

**8**(3–4), 231–357 (2015)Chambolle, A., Dossal, Ch.: On the convergence of the iterates of the “fast iterative shrinkage/thresholding algorithm”. J. Optim. Theory Appl.

**166**(3), 968–982 (2015)Du, S.S., Jin, C., Lee, J.D., Jordan, M.I., Singh, A., Poczos, B.: Gradient descent can take exponential time to escape saddle points. In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (eds.) Advances in Neural Information Processing Systems, vol. 30, pp. 1067–1077. Curran Associates Inc, Red Hook (2017)

Ghadimi, S., Lan, G.: Accelerated gradient methods for nonconvex nonlinear and stochastic programming. Math. Program.

**156**(1–2), 59–99 (2016)Jin, C., Netrapalli, P., Jordan, M.I.: Accelerated gradient descent escapes saddle points faster than gradient descent. arXiv preprint arXiv:1711.10456, (2017)

Lee, J.D., Simchowitz, M., Jordan, M.I., Recht, B.: Gradient descent only converges to minimizers. JMLR Workshop Conf. Proc.

**49**(1), 1–12 (2016)Li, H., Lin, Z.: Accelerated proximal gradient methods for nonconvex programming. In: Cortes, C., Lawrence, N.D., Lee, D.D., Sugiyama, M., Garnett, R. (eds.) Advances in Neural Information Processing Systems, vol. 28, pp. 379–387. Curran Associates Inc, Red Hook (2015)

Nesterov, Y.: A method for unconstrained convex problem with the rate of convergence \(O(1/k^2)\). Dokl AN SSSR

**269**, 543–547 (1983)Nesterov, Y.: Introductory Lectures on Convex Optimization: A Basic Course. Springer, New York (2004)

Polyak, B.T.: Introduction to Optimization. Optimization Software (1987)

Recht, B., Wright, S.J.: Nonlinear Optimization for Machine Learning (2017). (Manuscript in preparation)

Shub, M.: Global Stability of Dynamical Systems. Springer, Berlin (1987)

Tseng, P.: On accelerated proximal gradient methods for convex-concave optimization. Technical report, Department of Mathematics, University of Washington, (2008)

Zavriev, S.K., Kostyuk, F.V.: Heavy-ball method in nonconvex optimization problems. Comput. Math. Model.

**4**(4), 336–341 (1993)

## Acknowledgements

We are grateful to Bin Hu for his advice and suggestions on the manuscript. We are also grateful to the referees and editor for helpful suggestions.

## Author information

### Authors and Affiliations

### Corresponding author

## Additional information

This work was supported by NSF Awards IIS-1447449, 1628384, 1634597, and 1740707; AFOSR Award FA9550-13-1-0138; and Subcontract 3F-30222 from Argonne National Laboratory. Part of this work was done while the second author was visiting the Simons Institute for the Theory of Computing, and partially supported by the DIMACS/Simons Collaboration on Bridging Continuous and Discrete Optimization through NSF Award CCF-1740425.

## A properties of the sequence \(\{t_k\}\) defined by (51)

### A properties of the sequence \(\{t_k\}\) defined by (51)

In this “Appendix” we show that the following two properties hold for the sequence defined by (51):

and

We begin by noting two well known properties of the sequence \(t_k\) (see for example [4, Section 3.7.2]):

and

To prove that \(\frac{t_{k-1}-1}{t_k}\) is monotonically increasing, we need

Since \(t_{k+1} \ge t_k\) (which follows immediately from (51)), it is sufficient to prove that

By manipulating this expression and using (54), we obtain the equivalent expression

By definition of \(t_{k+1}\), we have

Thus (56) holds, so the claim (52) is proved. The sequence \(\{ (t_{k-1}-1)/t_k \}\) is nonnegative, since \((t_0-1)/t_1 = 0\).

Now we prove (53). We can lower-bound \((t_{k-1}-1)/t_k\) as follows:

For an upper bound, we have from \(t_k \ge t_{k-1}\) that

Since \(t_{k-1} \rightarrow \infty \) (because of (55)), it follows from (57) and (58) that (53) holds.

## Rights and permissions

## About this article

### Cite this article

O’Neill, M., Wright, S.J. Behavior of accelerated gradient methods near critical points of nonconvex functions.
*Math. Program.* **176**, 403–427 (2019). https://doi.org/10.1007/s10107-018-1340-y

Received:

Accepted:

Published:

Issue Date:

DOI: https://doi.org/10.1007/s10107-018-1340-y

### Keywords

- Accelerated gradient methods
- Nonconvex optimization

### Mathematics Subject Classification

- 90C26
- 49M30