English
Related papers

Related papers: First-order methods almost always avoid saddle poi…

200 papers

Is it possible for a first-order method, i.e., only first derivatives allowed, to be quadratically convergent? For univariate loss functions, the answer is yes -- the Steffensen method avoids second derivatives and is still quadratically…

Optimization and Control · Mathematics 2022-11-29 Minda Zhao , Zehua Lai , Lek-Heng Lim

The usual approach to developing and analyzing first-order methods for smooth convex optimization assumes that the gradient of the objective function is uniformly smooth with some Lipschitz constant $L$. However, in many settings the…

Optimization and Control · Mathematics 2017-10-11 Haihao Lu , Robert M. Freund , Yurii Nesterov

We propose a family of optimization methods that achieve linear convergence using first-order gradient information and constant step sizes on a class of convex functions much larger than the smooth and strongly convex ones. This larger…

Optimization and Control · Mathematics 2018-09-14 Chris J. Maddison , Daniel Paulin , Yee Whye Teh , Brendan O'Donoghue , Arnaud Doucet

This paper shows that a perturbed form of gradient descent converges to a second-order stationary point in a number iterations which depends only poly-logarithmically on dimension (i.e., it is almost "dimension-free"). The convergence rate…

Machine Learning · Computer Science 2017-03-03 Chi Jin , Rong Ge , Praneeth Netrapalli , Sham M. Kakade , Michael I. Jordan

Motivated by applications in Optimization, Game Theory, and the training of Generative Adversarial Networks, the convergence properties of first order methods in min-max problems have received extensive study. It has been recognized that…

Optimization and Control · Mathematics 2025-09-29 Constantinos Daskalakis , Ioannis Panageas

Randomly initialized first-order optimization algorithms are the method of choice for solving many high-dimensional nonconvex problems in machine learning, yet general theoretical guarantees cannot rule out convergence to critical points of…

Optimization and Control · Mathematics 2018-09-28 Dar Gilboa , Sam Buchanan , John Wright

It is known that gradient descent (GD) on a $C^2$ cost function generically avoids strict saddle points when using a small, constant step size. However, no such guarantee existed for GD with a line-search method. We provide one for a…

Optimization and Control · Mathematics 2025-12-17 Andreea-Alexandra Muşat , Nicolas Boumal

We give the first polynomial time algorithms for escaping from high-dimensional saddle points under a moderate number of constraints. Given gradient access to a smooth function $f \colon \mathbb R^d \to \mathbb R$ we show that (noisy)…

Machine Learning · Computer Science 2023-04-21 Dmitrii Avdiukhin , Grigory Yaroslavtsev

Many modern machine learning applications - from online principal component analysis to covariance matrix identification and dictionary learning - can be formulated as minimization problems on Riemannian manifolds, and are typically solved…

Optimization and Control · Mathematics 2023-11-07 Ya-Ping Hsieh , Mohammad Reza Karimi , Andreas Krause , Panayotis Mertikopoulos

We study the performance of stochastic first-order methods for finding saddle points of convex-concave functions. A notorious challenge faced by such methods is that the gradients can grow arbitrarily large during optimization, which may…

Machine Learning · Computer Science 2024-06-10 Gergely Neu , Nneka Okolo

First-order methods for solving convex optimization problems have been at the forefront of mathematical optimization in the last 20 years. The rapid development of this important class of algorithms is motivated by the success stories…

Optimization and Control · Mathematics 2021-01-07 Pavel Dvurechensky , Mathias Staudigl , Shimrit Shtern

We consider minimizing a nonconvex, smooth function $f$ on a Riemannian manifold $\mathcal{M}$. We show that a perturbed version of Riemannian gradient descent algorithm converges to a second-order stationary point (and hence is able to…

Optimization and Control · Mathematics 2019-06-19 Yue Sun , Nicolas Flammarion , Maryam Fazel

The convergence behaviour of first-order methods can be severely slowed down when applied to high-dimensional non-convex functions due to the presence of saddle points. If, additionally, the saddles are surrounded by large plateaus, it is…

Optimization and Control · Mathematics 2023-09-12 Nick Tsipinakis , Panos Parpas

Two classes of methods have been proposed for escaping from saddle points with one using the second-order information carried by the Hessian and the other adding the noise into the first-order information. The existing analysis for…

Optimization and Control · Mathematics 2018-03-05 Yi Xu , Rong Jin , Tianbao Yang

In centralized settings, it is well known that stochastic gradient descent (SGD) avoids saddle points and converges to local minima in nonconvex problems. However, similar guarantees are lacking for distributed first-order algorithms. The…

Optimization and Control · Mathematics 2022-03-07 Brian Swenson , Ryan Murray , H. Vincent Poor , Soummya Kar

Dynamical systems theory has recently been applied in optimization to prove that gradient descent algorithms bypass so-called strict saddle points of the loss function. However, in many modern machine learning applications, the required…

Machine Learning · Computer Science 2024-09-12 Patrick Cheridito , Arnulf Jentzen , Florian Rossmannek

This paper considers the problem of understanding the exit time for trajectories of gradient-related first-order methods from saddle neighborhoods under some initial boundary conditions. Given the 'flat' geometry around saddle points,…

Optimization and Control · Mathematics 2023-10-10 Rishabh Dixit , Mert Gurbuzbalaban , Waheed U. Bajwa

First-order algorithms have been popular for solving convex and non-convex optimization problems. A key assumption for the majority of these algorithms is that the gradient of the objective function is globally Lipschitz continuous, but…

Optimization and Control · Mathematics 2024-02-07 Junyu Zhang , Mingyi Hong

Gradient descent (GD) and stochastic gradient descent (SGD) are the workhorses of large-scale machine learning. While classical theory focused on analyzing the performance of these methods in convex optimization problems, the most notable…

Machine Learning · Computer Science 2019-09-05 Chi Jin , Praneeth Netrapalli , Rong Ge , Sham M. Kakade , Michael I. Jordan

Selecting an effective step-size is a fundamental challenge in first-order optimization, especially for problems with non-Euclidean geometries. This paper presents a novel adaptive step-size strategy for optimization algorithms that rely on…

Optimization and Control · Mathematics 2025-10-14 Abbas Khademi , Antonio Silveti-Falls