-Developed Bayesian Market Mixing Model for BEX GBV model-Enhanced Bayesian MMM model framework.-Developing the prediction process for the Bayesian Market Mixing Model with different choice of channel attribution.

Market Risk, SPG, Time Series

- Extracted, transformed, and loaded historical sales data w/o promotion events and utilized a hierarchical time series structure.- Built a demand forecast model on the open-to-buy (OTB) level using Fbprohet & DeepAR estimators with 50% accuracy improvement compared with the original rule-based model.

Team: Baidu Cognitive Research LabMentor: Dr. Ping Li–Solved PDEs by the neural networks and learned PDE-constrained inverse problems using these surrogate models.–Took research on the quantum Chernoff bound and proved the best lower bound.

Team: Mathematics & Computational Science Division Mentor: Dr. Wendy Di– Designed parallel computation codes with PETSC on Tomography– Derived different regularizers for Tomography problem with Center of Rotation

Team: Computational Sciences & Mathematics Mentor: Dr. Xiu Yang– Proved coefficients of n-th order ridge function can be approximated by n-sparsity vector under reasonable rotation matrix with O(\sqrt(n)) error– Design the Recurrent Neural Network(RNN) code to learn Wave Equation

Project Title: Optimal Control Under Uncertainty Supervisor: Prof. Georg Stadler– Derived optimality conditions of elliptic linear-quadratic optimal control problem– Derived Generalized Newton algorithm for the optimal control problem– Design the code for nonlinear situation

Project Title: Infinite Property of Local Structure of Compressive Sensing Supervisor: Prof. Ben Adcock– Expanded the local structure of Compressive Sensing to infinite situation– Calculated the sharpness of RIP of uniform recovery

Project Title: Nonuniform Sampling and Efficient Compressed Sensing MRI Supervisor: Prof. Benjamin James Stevens Adcock(SFU)– Realized Split Bregman method with signals’ and images’ denosing by Matlab– Compiled the Matlab program of Split Bregman method on MRI– Found the upper bound of the error between theoretical measurements and practical mea-surements– Proved the RIP in levels can be controlled by some constant in the constraints (Open Prob-lem).