Tianwei (Victor) Ni

I am a PhD student in Université de Montréal & Mila - Quebec AI Institute, advised by Pierre-Luc Bacon.

My research interests are reinforcement learning, or in a broad sense, the intersection of machine learning and control. I aim to understand the principles and connect the dots, design and implement practical algorithms, to solve important problems in this interdisciplinary field.

Prior to moving to Canada, I grew up and studied in China and USA. I obtained my Bachelor's degree in Computer Science at Peking University, and Master's degree in Machine Learning at Carnegie Mellon University.

Over the past few years, I was fortunate to collaborate with Jordi Salvador and Luca Weihs at AI2 on embodied AI, Ben Eysenbach and Russ Salakhutdinov at CMU on reinforcement learning, Katia Sycara at CMU on human-agent teaming, Lingxi Xie and Alan Yuille at JHU on medical image analysis.

Email  /  CV  /  Google Scholar  /  Semantic Scholar  /  Twitter  /  Github

profile photo
Selected Papers

Below are selected papers in reverse chronological order, and please see the full publication list in Google Scholar.
Notation: * indicates equal contribution, ° indicates equal advising.

Recurrent Model-Free RL is a Strong Baseline for Many POMDPs
Tianwei Ni, Benjamin Eysenbach, Ruslan Salakhutdinov
Preprint, 2021
project page / arXiv / code

Find and implement simple but strong baselines for POMDPs.

f-IRL: Inverse Reinforcement Learning via State Marginal Matching
Tianwei Ni*, Harshit Sikchi*, Yufei Wang*, Tejus Gupta*, Lisa Lee°, Ben Eysenbach°
CoRL, 2020
project page / arXiv / code

Learn state-only reward functions for downstream tasks.

Elastic Boundary Projection for 3D Medical Image Segmentation
Tianwei Ni, Lingxi Xie, Huangjie Zheng, Elliot K Fishman, Alan Yuille
CVPR, 2019
arXiv / code

Reformulate object segmentation as behavior cloning problem.

Service
teaching Graduate Teaching Assistant, 10-703 Deep Reinforcement Learning and Control, Carnegie Mellon University, Fall 2020

Last updated: Nov 27, 2021. Website template is credit to Jon Barron's source code.