Abstract:Despite its important applications in Machine Learning, min-max optimization of nonconvex-nonconcave objectives remains elusive. Not only are there no known first-order methods converging even to approximate local min-max points, but the computational complexity of identifying them is also poorly understood. In this paper, we provide a characterization of the computational complexity of the problem, as well as of the limitations of first-order methods in constrained min-max optimization problems with nonconvex-nonconcave objectives and linear constraints. As a warm-up, we show that, even when the objective is a Lipschitz and smooth differentiable function, deciding whether a min-max point exists, in fact even deciding whether an approximate min-max point exists, is NP-hard. More importantly, we show that an approximate local min-max point of large enough approximation is guaranteed to exist, but finding one such point is PPAD-complete. The same is true of computing an approximate fixed point of Gradient Descent/Ascent. An important byproduct of our proof is to establish an unconditional hardness result in the Nemirovsky-Yudin model. We show that, given oracle access to some function $f : P \to [-1, 1]$ and its gradient $\nabla f$, where $P \subseteq [0, 1]^d$ is a known convex polytope, every algorithm that finds a $\varepsilon$-approximate local min-max point needs to make a number of queries that is exponential in at least one of $1/\varepsilon$, $L$, $G$, or $d$, where $L$ and $G$ are respectively the smoothness and Lipschitzness of $f$ and $d$ is the dimension. This comes in sharp contrast to minimization problems, where finding approximate local minima in the same setting can be done with Projected Gradient Descent using $O(L/\varepsilon)$ many queries. Our result is the first to show an exponential separation between these two fundamental optimization problems.

On the complexity of finding a local minimizer of a quadratic function over a polytope

The Mixed Integer Trust Region Problem

The Complexity of Computing KKT Solutions of Quadratic Programs

Efficient Exact Algorithms for Minimum Covering of Orthogonal Polygons with Squares

The Complexity of Constrained Min-Max Optimization

On the Polyhedral Decision Problem

On the complexity of quasiconvex integer minimization problem

Hyperplanes Avoiding Problem and Integer Points Counting in Polyhedra

Complexity of Geometric programming in the Turing model and application to nonnegative tensors

The Nearest Complex Polynomial with a Zero in a Given Complex Domain

Spurious local minima in nonconvex sum-of-squares optimization

A Polynomial Algorithm for Some Relaxed Subset Sum Problems with Real Numbers

Computing local minimizers in polynomial optimization under genericity conditions

Complexity of Minimizing Regularized Convex Quadratic Functions

Integer Feasibility of Random Polytopes

On an Open Problem Related to the Strict Local Minima of Multilinear Objective Functions.

On the hardness of deciding the finite convergence of Lasserre hierarchies

Proving Nonnegativity of Polynomial with Computer

On the (Classical and Quantum) Fine-Grained Complexity of Log-Approximate CVP and Max-Cut

Sums of squares representations on singular loci

The Computational Complexity of Finding Stationary Points in Non-Convex Optimization