Active learning with generalized sliced inverse regression for high-dimensional reliability analysis

It is computationally expensive to predict reliability using physical models at the design stage if many random input variables exist. This work introduces a dimension reduction technique based on generalized sliced inverse regression (GSIR) mitigate curse of dimensionality. The proposed high dimensional method enables active learning integrate GSIR, Gaussian Process (GP) modeling, and Importance Sampling (IS), resulting in an accurate prediction reduced computational cost. new consists three core steps, 1) identification importance sampling region, 2) by GSIR produce sufficient predictor, 3) construction GP model for true response with respect predictor reduced-dimension space. High accuracy efficiency are achieved that iteratively executed above steps adding training points one region chance failure.