Yu Sun

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2023-86
May 10, 2023

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2023/EECS-2023-86.pdf

Most models in machine learning today are fixed during deployment. As a consequence, a trained model must prepare to be robust to all possible futures, even though only one of them is actually going to happen. The basic idea of test-time training is to train on this future once it arrives in the form of a test instance. Since each test instance arrives without a ground truth label, training is performed with self-supervision. This thesis explores the first steps in realizing this idea, for images, videos and robotics.

Advisor: Alexei (Alyosha) Efros and Moritz Hardt

BibTeX citation:

@phdthesis{Sun:EECS-2023-86,
    Author = {Sun, Yu},
    Title = {Test-Time Training},
    School = {EECS Department, University of California, Berkeley},
    Year = {2023},
    Month = {May},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2023/EECS-2023-86.html},
    Number = {UCB/EECS-2023-86},
    Abstract = {Most models in machine learning today are fixed during deployment. As a consequence, a trained model must prepare to be robust to all possible futures, even though only one of them is actually going to happen. The basic idea of test-time training is to train on this future once it arrives in the form of a test instance. Since each test instance arrives without a ground truth label, training is performed with self-supervision. This thesis explores the first steps in realizing this idea, for images, videos and robotics.}
}

EndNote citation:

%0 Thesis
%A Sun, Yu
%T Test-Time Training
%I EECS Department, University of California, Berkeley
%D 2023
%8 May 10
%@ UCB/EECS-2023-86
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2023/EECS-2023-86.html
%F Sun:EECS-2023-86