Comparison and Transfer Between Tasks in Overparameterized Learning
Offered By: Institute for Pure & Applied Mathematics (IPAM) via YouTube
Course Description
Overview
Explore a 49-minute lecture on "Comparison and transfer between tasks in overparameterized learning" presented by Vidya Muthukumar from the Georgia Institute of Technology at IPAM's Theory and Practice of Deep Learning Workshop. Delve into the similarities and differences between classification and regression tasks in overparameterized linear and kernel models. Discover high-dimensional regimes where regression fails to generalize while classification succeeds. Examine the concept of task transfer, focusing on using classification-trained model parameters for regression tasks. Learn about a post-processing algorithm that enables few-shot generalization in regression using classification-trained models. Gain insights into the fine-grained characterization of individual parameters resulting from minimum-norm interpolation on regression and classification tasks.
Syllabus
Vidya Muthukumar - Comparison and transfer between tasks in overparameterized learning
Taught by
Institute for Pure & Applied Mathematics (IPAM)
Related Courses
علم اجتماع المايكروباتKing Saud University via Rwaq (رواق) Statistical Learning with R
Stanford University via edX More Data Mining with Weka
University of Waikato via Independent The Caltech-JPL Summer School on Big Data Analytics
California Institute of Technology via Coursera Machine Learning for Musicians and Artists
Goldsmiths University of London via Kadenze