A Robust Asymmetric Kernel Function for Bayesian Optimization, With Application to Image Defect Detection in Manufacturing Systems

Some response surface functions in complex engineering systems are usually highly nonlinear, unformed, and expensive to evaluate. To tackle this challenge, Bayesian optimization (BO), which conducts sequential design via a posterior distribution over the objective function, is critical method used find global optimum of black-box functions. Kernel play an important role shaping estimated function. The widely kernel e.g., radial basis function (RBF), very vulnerable susceptible outliers; existence outliers causing its Gaussian process (GP) surrogate model be sporadic. In article, we propose robust asymmetric elastic net (AEN-RBF). Its validity as computational complexity evaluated. When compared with baseline RBF kernel, prove theoretically that AEN-RBF can realize smaller mean squared prediction error under mild conditions. proposed also faster convergence optimum. We show less sensitive outliers, hence improves robustness corresponding BO GPs. Through extensive evaluations carried out on synthetic real-world problems, outperforms existing benchmark Note Practitioners—Some industrial cannot accurately represented by physical models. situation, data-driven necessary for advancing system automation intelligence. one strategies learning has been applied robotics, anomaly detection, automatic algorithm configuration, reinforcement learning, deep learning. This article proposes new named after AEN-RBF. will make GPs more lower data quality barrier training. was motivated hyperparameter tuning problem models image defect detection advanced manufacturing, but easily extended other applications where needed. Our verified problems.