Implementation of Zeroth-order Adaptive Perturbation (ZAP)

Code implementation of the ZAP algorithm for zeroth-order Gaussian Process (GP) hyperparameter optimization.

Usage

Install dependencies using uv:

# Install dependencies
uv sync

Key Parameters

• a, b: Initial learning rate and perturbation magnitude
• tau, gamma: Decay rates for learning rate and perturbation
• parallel: Enable parallel function evaluation
• averaging: Use gradient averaging for variance reduction
• q: Number of gradient estimates to average

Citation

If you use this code in your research, please cite our corresponding paper.