Student Seminar Series: Dr. Zhi (Kevin) He, Gateaux differentials based boosting for fitting large-scale survival data

Large-scale time to event data arise rapidly in biomedical studies. Data set are often extremely large in terms of number of observations and/or number of variables. Moreover, large-scale time event data are usually complex. The computational burden of model fitting increases quickly as the number of sample size or the number of predictors grows, which prohibits the application of exiting statistical methods.  We propose a new Gateaux differential-based boosting procedure for variable selection and prediction. The proposed method is built upon functional differentials and is applicable to complex settings including: 1. selection of time-varying effects in survival analysis; 2. dynamic detection of gene-gene interactions.