Chelsea Finn - Principles for Tackling Distribution Shift: Pessimism, Adaptation, and Anticipation
Abstract: While we have seen immense progress in machine learning, a critical shortcoming of current methods lies in handling distribution shift between training and deployment. Distribution shift is pervasive in real-world problems ranging from natural variation in the distribution over locations or domains, to shift in the distribution arising from different decision making policies, to shifts over time as the world changes. In this talk, I’ll discuss three general principles for tackling these forms of distribution shift: pessimism, adaptation, and anticipation. I’ll present the most general form of each principle before providing concrete instantiations of using each in practice. This will include a simple method for substantially improving robustness to spurious correlations, a framework for quickly adapting a model to a new user or domain with only unlabeled data, and an algorithm that enables robots to anticipate and adapt to shifts caused by other agents.
Bio: Chelsea Finn is an Assistant Profe
15 views
31
7
12 months ago 00:00:47 6
HumanPlus: Autonomous Skills from Imitating Humans
1 year ago 00:01:12 6
FF Live is BACK at Founders Forum Global 2024
1 year ago 00:02:38 1
Mobile ALOHA: Your Housekeeping Robot
1 year ago 00:09:18 1
Joe Cole [Best Skills & Goals]
1 year ago 00:01:40 7
Mobile ALOHA - A Smart Home Robot - Compilation of Autonomous Skills
1 year ago 00:00:55 1
Mobile ALOHA funny failures
2 years ago 00:03:29 3
FNF - Pibby Apocalypse - Come Along With Me (by Awe) - [FC/4k]
2 years ago 00:12:45 1
Life is Plastic
2 years ago 00:02:26 1
Vestri the robot imagines how to perform tasks
2 years ago 00:00:56 1
What inspired you to choose AI and Robotics as a profession? - Chelsea Finn & Andrew Ng
2 years ago 00:02:59 3
Nubiyan Twist - Tittle Tattle feat Cherise
2 years ago 00:04:36 2
Dead Cool - Stranger Kind (Official Video)
3 years ago 01:26:18 2
CS330: Deep Multi-task and Meta Learning | 2020 | Lecture 8 - Bayesian Meta-Learning
3 years ago 00:03:12 14
Looking Forward: AI in 2022
4 years ago 03:38:45 4
ICCV 2021 Workshop - 2nd Visual Inductive Priors for Data-Efficient Deep Learning Workshop
4 years ago 00:09:20 4
True Confessions with Chelsea Handler and Finn Wolfhard | The Tonight Show Starring Jimmy Fallon
4 years ago 06:27:03 1
Conceptual Understanding of Deep Learning Workshop
4 years ago 01:06:41 1
Principles for Tackling Distribution Shift: Pessimism, Adaptation, and Anticipation
4 years ago 00:48:57 15
Chelsea Finn - Principles for Tackling Distribution Shift: Pessimism, Adaptation, and Anticipation
4 years ago 00:01:33 1
England Beat Belgium (1952)
5 years ago 01:16:43 16
DLRL Summer School 2020 - Meta Reinforcement Learning - Chelsea Finn