Hi! I am Garrett Thomas, a PhD candidate in the Computer Science Department at Stanford, where I am fortunate to advised by Tengyu Ma. My academic interests lie broadly in reinforcement learning, in particular (deep) model-based approaches.

Previously I studied Computer Science and Mathematics at UC Berkeley, where I was part of Pieter Abbeel's group, working with Aviv Tamar on reinforcement learning and its applications to robotics. Our research was largely concerned with policy structure, planning, and generalization.

You can reach me at gwthomas@stanford.edu.

PLEX: Making the Most of the Available Data for Robotic Manipulation Pretraining
Garrett Thomas, Ching-An Cheng, Ricky Loynd, Felipe Vieira Frujeri, Vibhav Vineet, Mihai Jalobeanu, Andrey Kolobov
CoRL 2023
paper

Safe Reinforcement Learning by Imagining the Near Future
Garrett Thomas, Yuping Luo, Tengyu Ma
NeurIPS 2021
paper code

MOPO: Model-based Offline Policy Optimization
Tianhe Yu*, Garrett Thomas*, Lantao Yu, Stefano Ermon, James Zou, Sergey Levine, Chelsea Finn†, Tengyu Ma†
NeurIPS 2020
paper code

Model-based Adversarial Meta-Reinforcement Learning
Zichuan Lin, Garrett Thomas, Guangwen Yang, Tengyu Ma
NeurIPS 2020
paper code

Learning Robotic Assembly from CAD
Garrett Thomas*, Melissa Chien*, Aviv Tamar, Juan Aparicio Ojea, Pieter Abbeel
ICRA 2018, Finalist for Automation Award
paper code

Learning from the Hindsight Plan – Episodic MPC Improvement
Aviv Tamar, Garrett Thomas, Tianhao Zhang, Sergey Levine, Pieter Abbeel
ICRA 2017, Winner of Best Poster Award at NIPS 2016 Workshop on Neurorobotics
paper

Value Iteration Networks
Aviv Tamar, Yi Wu, Garrett Thomas, Sergey Levine, Pieter Abbeel
NIPS 2016, Winner of Best Paper Award
paper code

Force

I am developing this library to support my research.

Notes

While an undergraduate at Berkeley, I was heavily involved in the production of notes for CS 189, Introduction to Machine Learning. In particular, I wrote a summary of math background needed for the course, and then co-authored notes for the course itself. I am no longer involved with CS 189, but these materials appear to be still in use.

I found this experience extremely valuable and am trying to continue writing notes in my (now more limited) spare time.

illume

A number of years ago (in high school) I developed a light-based puzzle game called illume. It used to be available on the iOS App Store, but I decided it wasn't worth paying for Apple's developer program to continue offering a free app. Prior to its removal it was downloaded over 22,000 times and received many positive reviews from users. A video of the gameplay can be viewed here.