DEC 6th - DEC 14th

NeurIPS 2021

Training virtual robots in 3D environments. Discovering new catalysts for renewable energy storage. Training speech recognition without any labeled data. These are some of the topics that Meta AI researchers will present new work on at the virtual Conference on Neural Information Processing Systems (NeurIPS), one of the largest AI conferences of the year. Our researchers and technologists work closely with open-source communities, academia, and partners to build the future of AI together.

Our Work

Zero-Shot Transfer to Formula 1 Race Tracks

This shows agents trained using our methods (PLR and REPAIRED) outperforming baseline agents in zero-shot transfer to the challenging task of driving a track in the shape of the Nürburging Grand Prix racetrack. This means they are able to drive this track without having ever trained on it.

See Demo

Building diverse MiniHack environments within just a few lines of code

MiniHack is a sandbox framework for easily designing rich and diverse environments for Reinforcement Learning. To this end, MiniHack leverages description files written using a human-readable domain-specific language. With just a few lines of code, people can generate a large variety of Gym environments, ranging from small rooms to complex, procedurally generated worlds.

See Demo

Habitat 2.0: Training Home Assistants to Rearrange their Habitat

The Habitat 2.0 simulation platform is for training virtual robots in interactive, physics-enabled, 3D environments. We present the new ReplicaCAD 3D dataset of apartments, the Habitat 2.0 simulator capable of simulating 850x faster than real-time, and the Home Assistant Benchmark, a suite of common

See Demo

Featured Publications

Unsupervised Speech Recognition

Wed Dec 8, 8:00 AM - 8:15 AM PT

Alexei Baevski, Wei-Ning Hsu, Alexis Conneau, Michael Auli

Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers

Wed Dec 8, 8:20 AM - 8:35 AM PT

Mandela Patrick, Dylan Campbell, Yuki Asano, Ishan Misra, Florian Metze, Christoph Feichtenhofer, Andrea Vedaldi, Joao Henriques

Volume Rendering of Neural Implicit Surfaces

Wed Dec 8, 8:40 AM - 8:55 AM PT

Lior Yariv, Jiatao Gu, Yoni Kasten, Yaron Lipman

Interesting Object, Curious Agent: Learning Task-Agnostic Exploration

Fri Dec 10, 4:00 PM - 4:15 PM PT

Simone Parisi, Victoria Dean, Deepak Pathak, Abhinav Gupta

Moser Flow: Divergence-based Generative Modeling on Manifolds

Fri Dec 10, 4:00 PM - 4:15 PM PT

Noam Rozen, Aditya Grover, Maximilian Nickel, Yaron Lipman

See All Papers

Sponsored Workshops

WiML

The 16th Workshop for Women in Machine Learning co-located with NeurIPS will feature invited speakers, oral and poster presentations, and more.

LatinX in AI

One-day workshop will feature invited speakers, oral presentations, poster sessions, panel discussion and a mentoring session.

Black in AI

5th Black in AI Workshop co-located @NeurIPS 2021 will feature invited talks from prominent researchers and practitioners and a poster session.

Invited Talks

Workshops

Spotlight Papers

Dec 7, Poster session 2

Yann Dubois, Benjamin Bloem-Reddy, Karen Ullrich, Chris Maddison

Multimodal and Multilingual Embeddings for Large-Scale Speech Mining

Dec 8, Poster session 3

Paul-Ambroise Duquenne, Hongyo Gong, Holger Schwenk

Habitat 2.0: Training Home Assistants to Rearrange their Habitat

Dec 9, Poster session 5

Andrew Szot, Alexander Clegg, Eric Undersander, Erik Wijmans, Yili Zhao, John Turner, Noah Maestre, Mustafa Mukadam, Devendra Singh Chaplot, Oleksandr Maksymets, Aaron Gokaslan, Vladimir Vondrus, Sameer Dharur, Franziska Meier, Wojciech Galuba, Angel Chang, Zsolt Kira, Vladlen Koltun, Jitendra Malik, Manolis Savva, Dhruv Batra

A Provably Efficient Sample Collection Strategy for Reinforcement Learning