欢迎光临!
您现在所在的位置:首页 >> 通知公告 & 学术信息
学术信息
SEMINARS
Deep Learning, Prediction and Validation: Modelling and Analysis of Complex Data in Technology in Modern Science and Technology
时间  Datetime
2018-12-14 14:00 — 15:00 
地点  Venue
Large Conference Room
报告人  Speaker
Tze Leung Lai
单位  Affiliation
Stanford University
邀请人  Host
Dong Han
报告摘要  Abstract

Concerning "prediction" and "modelling" in the title, we begin with a review of (a) recent advances in computer vision and deep learning, (b) the underlying mathematical theory of convolutional neural networks, gradient descent, and hidden Markov random fields, and (c) AI applications to medical imaging and fast automated analysis of strong gravitational lenses in astrophysics. Whereas high-performance computing and advanced programming have overcome the computational hurdles in the analysis of "big data" in modern science and technology, we show how novel mathematical methods and statistical principles can provide major breakthroughs in the "validation" of scientific theories based on complex experimental data. Because big data typically require variable/hypothesis selection based on some sparsity assumption to make the inference problem feasible, there is contemporaneous awareness of how this complexity has led to irreproducible research in modern science. We describe statistical innovations in postselection inference and hybrid resampling to address this "reproducible (replication) crisis", and illustrate our point with the care in collecting data and their analysis in the Higgs boson experiments.