Mixed-Sample SGD: an End-to-end Analysis of Supervised Transfer Learning

Yuyang Deng (Columbia University) · Samory Kpotufe (Columbia University)

adaptive sub-samplingconstrained convex programsconvergence rate convex losses empirical validation linear regressionmixed-sample sgdnegative transfer optimization techniques prediction taskssgd proceduresource data qualitysquare loss statistical performancestatistical transfer guaranteessupervised transfer learning

Theoretical works on supervised transfer learning (STL)