Xing Shen

I leaded a research project on advancing robustness and uncertainty estimates in medical image analysis under sample quality variations, and produced a research paper as the first author; Further more, I adapted the precision and recall metric for 3D brain MRI image synthesis with a focus on multiple sclerosis investigation from a NeurIPS 2019 paper: "Improved Precision and Recall Metric for Assessing Generative Models", and it helped other ongoing projects in the lab to access model performance.

I participated in formulation of research in drug response prediction and its methodologies which utilizing multi-modal cancer cell line features and drug descriptors to predict response sensitivity. Collected data and conducted experiments for transfer learning across two major datasets. And developed novel end-to-end deep learning model based on transformers to predict drug response. The implementation is available at: https://github.com/xingbpshen/MTDRP.

I leaded a research on visual reasoning based on neural-symbolic methods to fill in the gap between differentiable symbolic system and neural component for visual recognition. I scheduled project progress and held weekly group meetings. I developed a 2-stage pipeline with an image-symbol neural mapper and a tuned symbolic solver for tackling NP-complete problems in a supervised learning paradigm. And I produced a technical report as the first author. Code is available at: https://github.com/xingbpshen/SATNet.