Optimization on black-box function by parameter-shift rule

2025-03-16

AI Summary

This work addresses the high query cost and strong parameter dependence inherent in black-box function optimization. We propose a novel zeroth-order optimization method grounded in the quantum parameter-shift rule (PSR), marking the first integration of PSR into classical black-box optimization frameworks. Unlike conventional zeroth-order methods, our approach requires no gradient information and constructs accurate directional estimates using only a small number of function queries—bypassing the need for high query frequency or large parameter counts. Evaluated across diverse non-convex black-box tasks, the method achieves significantly faster convergence, reducing average query counts by approximately 40% and demonstrating superior parameter efficiency compared to state-of-the-art zeroth-order optimizers such as NES and ZOO. The proposed framework establishes a scalable, resource-efficient paradigm for black-box optimization under stringent computational constraints.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Optimization
• Reasoning under Uncertainty: Stochastic Optimization

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Security and Privacy: Large-scale security measurements

Abstract

Machine learning has been widely applied in many aspects, but training a machine learning model is increasingly difficult. There are more optimization problems named"black-box"where the relationship between model parameters and outcomes is uncertain or complex to trace. Currently, optimizing black-box models that need a large number of query observations and parameters becomes difficult. To overcome the drawbacks of the existing algorithms, in this study, we propose a zeroth-order method that originally came from quantum computing called the parameter-shift rule, which has used a lesser number of parameters than previous methods.

Problem

• Optimize black-box functions with uncertain parameter-outcome relationships.
• Reduce query observations and parameters in black-box model optimization.
• Introduce parameter-shift rule from quantum computing for efficient optimization.

Innovation

• Uses parameter-shift rule from quantum computing
• Reduces number of required parameters significantly
• Optimizes black-box functions more efficiently