Submodular optimization hits barrier in balancing quality and stability

A Sharp Barrier for Consistent Submodular Maximization: Any Improvement over $2-\sqrt{2}$ Entails Exponential Queries or Linear Recourse

Data Structures and AlgorithmsMachine Learning

Summary

The paper studies how algorithms can pick good sets of items when those items arrive over time, balancing between getting a high-quality choice and not changing the chosen set too much after each new arrival. The authors prove a sharp limit on how good the choices can be if the algorithm must stay stable and efficient: no efficient method can beat about 58.6% of the best possible offline choice. They also show that trying to improve beyond this limit either needs a huge number of checks or big changes to the chosen set. This helps explain the cost of keeping solutions stable when the future is uncertain.

What this means in practice

  • For online recommendation teams: Design recommendation algorithms that balance good item selection with minimal updates as new data arrives.
  • For supply chain planners: Develop selection methods for inventory or resources that remain stable despite frequent updates, improving operational consistency.

A theory result. No direct application yet.

Authors

Shi Fu, Qixin Zhang, Dacheng Tao

Abstract

Consistent submodular maximization studies the tradeoff between solution quality and stability when elements arrive over time. For a monotone submodular objective, which models diminishing returns, an algorithm maintains a set of at most $k$ available elements and changes only $O(1)$ elements after each insertion. Dütting et al. [2025] established a tight $2/3$ approximation with unrestricted computation and a polynomial-time $0.51$ approximation. They left open at STOC 2025 whether efficient algorithms can match the offline $1-1/e$ guarantee. We resolve this problem by proving that the supremum approximation achievable with polynomially many value queries and worst-case constant recourse is \[ β=2-\sqrt2\approx0.5858<1-1/e. \] For every $\varepsilon>0$, our randomized algorithm attains $β-\varepsilon$ with $O(\varepsilon^{-2})$ changes per insertion. Any fixed improvement requires exponentially many queries before one critical insertion or linear recourse of $Ω(k)$ changes at that insertion, even with unlimited queries afterwards. This gap quantifies the cost of consistency: the current oracle hides which elements will be needed after an arrival. We also determine the exact curvature-dependent threshold $1-(\sqrt2-1)\vartheta$, attain $1-1/e-\varepsilon$ for weighted coverage with $O(\varepsilon^{-1})$ recourse, and separate the existence of universal future-price certificates from their efficient computation. Our algorithm has a bounded-bit polynomial-time implementation for polynomial-bit rational oracle answers; the lower bound uses only logarithmic-bit rational answers.