Joint Artificial Intelligence Seminar / Computer Science Speaking Skills Talk

— 1:00pm

In Person - ASA Conference Room, Gates Hillman 6115

LUCIO DERY , Ph.D. Student, Computer Science Department, Carnegie Mellon University

An automated transfer learning approach to tackling learning under limited data

Transfer learning is arguably the engine of the current deep learning revolution in machine learning.  A common branch of transfer learning is learning with auxiliary objectives — supplementary learning signals that are introduced to help aid learning on data-starved or highly complex end-tasks. Whilst much work has been done to formulate useful auxiliary objectives, their construction is still an art which proceeds by slow and tedious hand-design. Intuition for how and when these objectives improve end-task performance has also had limited theoretical backing. 

In this talk, I will present a task agnostic approach for automatically generating a suite of auxiliary objectives and maximally utilizing them to benefit a specified end-task. We achieve this by deconstructing existing objectives within a novel unified taxonomy, identifying connections between them, and generating new ones based on the uncovered structure.  We theoretically formalize widely-held intuitions about how auxiliary learning improves generalization on the end-task which leads us to a principled and efficient algorithm for searching the space of generated objectives to find those most useful to a specified end-task.

Presented as part of the AI Seminar Series

Presented in Partial Fulfillment of the CSD Speaking Skills Requirement.

In Person and Zoom Participation. See announcement.

Event Website:

Add event to Google
Add event to iCal