Bayesian optimization method improves tuning under noisy conditions
Robust Bayesian Optimization with Q-Exponential Surrogates
Machine Learning
Summary
Optimizing complicated systems can be difficult when the data has unexpected errors or noise. The authors developed a new method called q-ED-BO that uses a flexible statistical model able to handle unusual or extreme data better than traditional methods. This approach adjusts its assumptions to manage noise and outliers without losing efficiency. Their tests showed it performs as well or better on clean data and significantly better when data is messy, especially in tuning audio processing devices.
What this means in practice
- •For signal processing engineers: Improve tuning of beamformers and adaptive filters in noisy environments by using a robust optimization method less sensitive to outliers.
- •For control system designers: Enhance robustness of parameter optimization for black-box systems affected by irregular measurement errors or interference.
Authors
Richard Cornelius Suwandi, Zhidi Lin, Feng Yin, Abdelhak M. Zoubir
Abstract
Bayesian optimization (BO) is a widely used framework for optimizing expensive black-box objectives, but standard BO methods often use Gaussian process (GP) surrogates whose Gaussian assumption is sensitive to outliers and heavy-tailed noise. We introduce q-ED-BO, a robust BO method whose surrogate follows a univariate q-exponential (q-ED) distribution, preserving GP-BO's closed-form posterior mean and variance while a shape parameter q controls the tail behavior, recovering the GP at q = 2 and growing heavier-tailed with wider confidence bounds as q decreases. This tractability yields a closed-form q-upper confidence bound (q-UCB) with sublinear regret, and an exact closed-form q-expected improvement (q-EI) that generalizes EI to the heavy-tailed predictive, recovering classical EI at q = 2. Experiments on beamformer and adaptive filter tuning with impulsive outliers show that q-ED-BO matches or exceeds existing baselines on clean data, and under corruption, improves the strongest baseline by approximately 0.7 dB in output SINR and 1.1 to 1.2 dB in misalignment reduction.