On the Computational Complexity of Performative Prediction
Preprint
Abstract
Performative prediction captures the phenomenon where deploying a predictive model shifts the underlying data distribution. While simple retraining dynamics are known to converge linearly when the performative effects are weak ($ρ< 1$), the complexity in the regime $ρ> 1$ was hitherto open. In this paper, we establish a sharp phase transition: computing an $ε$-performatively stable point is PPAD-complete -- and thus polynomial-time equivalent to Nash equilibria in general-sum games -- even when $ρ= 1 + O(ε)$. This intractability persists even in the ostensibly simple setting with a quadratic loss function and linear distribution shifts. One of our key technical contributions is to extend this PPAD-hardness result to general convex domains, which is of broader interest in the complexity of variational inequalities. Finally, we address the special case of strategic classification, showing that computing a strategic local optimum is PLS-hard.
Keywords
All Related Versions
This publication has 12 references indexed in Scilit:
- Pure-Circuit: Tight Inapproximability for PPADJournal of the ACM, 2024
- Multi-agent Performative Prediction: From Global Stability and Optimality to Chaos2023
- Settling the complexity of computing two-player Nash equilibriaJournal of the ACM, 2009
- The Complexity of Computing a Nash EquilibriumSIAM Journal on Computing, 2009
- Inexact Variants of the Proximal Point Algorithm without MonotonicitySIAM Journal on Optimization, 2003
- Approximating Fixed Points of Weakly Contracting MappingsJournal of Complexity, 1999
- On the complexity of the parity argument and other inefficient proofs of existenceJournal of Computer and System Sciences, 1994
- An Ellipsoid Algorithm for the Computation of Fixed PointsJournal of Complexity, 1993
- Approximation of fixed points of nonexpansive mappingsArchiv der Mathematik, 1992
- Exponential lower bounds for finding Brouwer fix pointsJournal of Complexity, 1989