Seminar by Prof. Zhonghua Liu from Biostatistics, Mailman School of Public Health, Columbia University

Positive synergies between statistics and deep learning harness their complementary strengths to tackle complex problems. In the first part of this talk, I will demonstrate how deep learning enhances causal inference within the nonparametric efficient influence function framework, with specific applications to algorithmic fairness. I will then introduce a novel approach that formulates semiparametric estimation as a bi-level optimization problem, leading to the development of deep neural-nets assisted semiparametric estimation (DNA-SE)—a scalable algorithm that leverages the universal approximation power of deep neural networks to streamline semiparametric procedures. In the second part, I will explore, from a traditional statistical perspective, why over-parameterized deep neural networks do not exhibit overfitting in practice.

Prof. Zhonghua Liu currently is an Assistant Professor of Biostatistics, Mailman School of Public Health, Columbia University. He obtained PhD degree in Biostatistics from Harvard University, and his current research interests include Causal Inference, Machine Learning, Statistical Genetics and Genomics.