Fairness-Regularized Online Optimization with Switching Costs

Agentic AI
Published: arXiv: 2512.11131v1
Authors

Pengfei Li Yuelin Han Adam Wierman Shaolei Ren

Abstract

Fairness and action smoothness are two crucial considerations in many online optimization problems, but they have yet to be addressed simultaneously. In this paper, we study a new and challenging setting of fairness-regularized smoothed online convex optimization with switching costs. First, to highlight the fundamental challenges introduced by the long-term fairness regularizer evaluated based on the entire sequence of actions, we prove that even without switching costs, no online algorithms can possibly achieve a sublinear regret or finite competitive ratio compared to the offline optimal algorithm as the problem episode length $T$ increases. Then, we propose FairOBD (Fairness-regularized Online Balanced Descent), which reconciles the tension between minimizing the hitting cost, switching cost, and fairness cost. Concretely, FairOBD decomposes the long-term fairness cost into a sequence of online costs by introducing an auxiliary variable and then leverages the auxiliary variable to regularize the online actions for fair outcomes. Based on a new approach to account for switching costs, we prove that FairOBD offers a worst-case asymptotic competitive ratio against a novel benchmark -- the optimal offline algorithm with parameterized constraints -- by considering $T\to\infty$. Finally, we run trace-driven experiments of dynamic computing resource provisioning for socially responsible AI inference to empirically evaluate FairOBD, showing that FairOBD can effectively reduce the total fairness-regularized cost and better promote fair outcomes compared to existing baseline solutions.

Paper Summary

Problem
The main problem this research paper addresses is the challenge of designing online optimization algorithms that balance two crucial considerations: fairness and action smoothness. In many online optimization problems, such as energy production scheduling, object tracking, or server capacity provisioning, it's essential to reduce abrupt changes in actions while ensuring fairness in outcomes.
Key Innovation
The innovation of this work is the proposal of FairOBD (Fairness-regularized Online Balanced Descent), a new algorithm that reconciles the tension between minimizing three costs: hitting cost, switching cost, and fairness cost. FairOBD decomposes the long-term fairness cost into a sequence of online costs by introducing an auxiliary variable and then leverages this variable to regularize online actions for fair outcomes.
Practical Impact
This research has significant practical implications for designing online optimization algorithms that promote fairness and action smoothness. The proposed FairOBD algorithm can be applied in various domains, such as dynamic computing resource provisioning for socially responsible AI inference, to reduce the total fairness-regularized cost and promote fair outcomes. By addressing fairness and action smoothness simultaneously, this research can help mitigate societal biases and promote socioeconomic fairness.
Analogy / Intuitive Explanation
Imagine you're designing a traffic light system to optimize traffic flow while ensuring fairness among all drivers. The traffic light should change smoothly to avoid sudden stops, but it should also ensure that all drivers have an equal chance to pass through the intersection. FairOBD is like a smart traffic light system that balances these two competing goals: smooth traffic flow and fairness among drivers. By doing so, it can optimize the overall traffic flow while promoting fairness and reducing congestion.
Paper Information
Categories:
cs.LG cs.AI
Published Date:

arXiv ID:

2512.11131v1

Quick Actions