DeepAI AI Chat
Log In Sign Up

A polynomial time log barrier method for problems with nonconvex constraints

by   Oliver Hinder, et al.

Interior point methods (IPMs) that handle nonconvex constraints such as IPOPT, KNITRO and LOQO have had enormous practical success. Unfortunately, all known analyses of log barrier methods with general constraints (implicitly) prove guarantees with exponential dependencies on 1/μ where μ is the barrier penalty parameter. This paper provides an IPM that finds a μ-approximate Fritz John point in O( μ^-7/4) iterations when the objective and constraints have Lipschitz first and second derivatives. For this setup, the results represent both the first polynomial time dependence on 1/μ for a log barrier method and the best-known guarantee for finding Fritz John points. We also show that, given convexity and regularity conditions, our algorithm finds an ϵ-optimal point in at most O(ϵ^-2/3) iterations. The algorithm that we study in this paper, although naive, provides inspiration for our practical one-phase IPM.


page 1

page 2

page 3

page 4


No self-concordant barrier interior point method is strongly polynomial

It is an open question to determine if the theory of self-concordant bar...

A Newton-CG based barrier-augmented Lagrangian method for general nonconvex conic optimization

In this paper we consider finding an approximate second-order stationary...

Polynomial time guarantees for the Burer-Monteiro method

The Burer-Monteiro method is one of the most widely used techniques for ...

Escape saddle points by a simple gradient-descent based algorithm

Escaping saddle points is a central research topic in nonconvex optimiza...

An infeasible interior-point arc-search method with Nesterov's restarting strategy for linear programming problems

An arc-search interior-point method is a type of interior-point methods ...

The order barrier for the L^1-approximation of the log-Heston SDE at a single point

We study the L^1-approximation of the log-Heston SDE at the terminal tim...