报告题目：A low rank approximation and its applications in uncertainty quantification
摘要：A low rank approximation is presented for efficient real-time computation of stochastic models. In the approach, a novel variable-separation is used to get a separated representation of the solution for stochastic models in a systematic enrichment manner. A model-driven stochastic basis functions are constructed in the low rank approximation. To significantly decrease the computation complexity for the stochastic basis functions, we construct a hybrid low rank approximation based on multi-fidelity models and multiple models. The proposed approach is explored in uncertainty quantification, e.g., stochastic saddle point problems, Bayesian inversion and data assimilation.