Neural Information Processing Systems (NeurIPS) is an annual interdisciplinary conference focusing on machine learning, artificial intelligence (AI), and statistics. This year, the 36th NeurIPS conference was a hybrid event, with researchers gathering in person at the New Orleans Convention Center during the first week and attending virtually during the second.
ServiceNow participated in various ways: with a scientific contribution of four papers accepted at the main conference and 11 papers accepted at workshops, members of the research group on-site to represent the company, a Gold sponsorship booth, and the organization of a joint ServiceNow/Hugging Face event associated with the BigCode project.
Continue reading for highlights from one of the most anticipated scientific events of the year and details about the specific contributions of ServiceNow Research.
The NeurIPS conference took place Nov. 29 to Dec. 9, 2022, with researchers from around the world gathering at the New Orleans Convention Center during the first week. Throughout this multitrack interdisciplinary meeting focusing on machine learning, AI, and statistics, researchers engaged with their peers during invited talks, oral and poster presentations, demonstrations, competitions, tutorials, and workshops.
One of the keynote speakers, Geoff Hinton, a faculty member at the University of Toronto, presented a talk titled “The Forward-Forward Algorithm for Training Deep Neural Networks” that was particularly appreciated by ServiceNow Research Vice President Nicolas Chapados.
Chapados noted, “In a flashback to 2006, Hinton proposes to replace backprop with greedy layerwise training of deep networks using two forward passes instead. Importantly, this approach provides a nice way to incorporate both high-level context and low-level perception in predictions—and might be related to why we sleep."
As part of the 2022 NeurIPS conference, ServiceNow Research members contributed to 15 papers. See the full list, including authors and links to papers, below. Members of the ServiceNow Research team were on-site to represent the company. According to Research Scientist Valentina Zantedeschi, “Going to NeurIPS is always an enriching and fun experience. This year too I was exposed to new ideas, and I had the chance to discuss extensions of my research works and the big picture of where AI is headed.”
As a Gold sponsor, ServiceNow had a booth in the exhibitor hall where attendees interested in the fundamental and applied work conducted by the ServiceNow Research group could ask questions, look at research demos, and chat with researchers. Several researchers even showcased their scientific contributions during individual poster sessions, which generated a lot of interest and excitement.
David Vazquez, research scientist and manager of research programs, spent most of his time at the booth engaging with researchers from a variety of fields. “This year was the first time ServiceNow Research had a booth at a machine learning conference,” he said. “It was refreshing to talk to many NeurIPS attendees about their research and job opportunities at ServiceNow.”
Toward the end of the in-person portion of NeurIPS 2022, the ServiceNow team hosted a networking event in collaboration with Hugging Face at Juan’s Flying Burrito. More than 80 members of the research community came together for lively conversations about the BigCode project and its positive impact on the responsible development of large language models for code, as well as how BigCode empowers developers through open governance.
We also had a fun quiz about The Stack. This was one of the most appreciated social gatherings of the conference. Thank you to everyone who participated.
ServiceNow Research’s involvement in NeurIPS was a success on all fronts. Congratulations to the organizers, volunteers, and participants!
On Margins and Generalisation for Voting Classifiers (Felix Biggs, Valentina Zantedeschi, Benjamin Guedj)
Neural Attentive Circuits (Martin Weiss, Yoshua Bengio, Nassim, Bernhard Schölkopf, Christopher Pal)
Masked Conditional Video Diffusion for Prediction, Generation, and Interpolation (Vikram Voleti, Alexia Jolicoeur-Martineau, Christopher Pal)
Attention-based Neural Cellular Automata (Mattie Tesfaldet, Derek Nowrouzezahrai, Christopher Pal)
NeurIPS workshop papers
Using Confounded Data in Offline RL (Maxime Gasse, Damien Grasset, Guillaume Gaudron, Pierre-Yves Oudeyer)
Implicit Offline Reinforcement Learning via Supervised Learning (Alexandre Piché, Rafael Pardinas, David Vazquez, Igor Mordatch, Christopher Pal)
Exploring the Design Space of Generative Diffusion Processes for Sparse Graphs (Pierre-André Noël, Pau Rodriguez)
Discrete Learning Of DAGs Via Backpropagation (Andrew J. Wren, Pasquale Minervini, Luca Franceschi, Valentina Zantedeschi)
Contrastive Self-supervision Defines General-Purpose Similarity Functions (Charles Guille-Escuret, Pau Rodriguez, David Vazquez, Ioannis Mitliagkas, João Monteiro)
Constraining Low-level Representations to Define Effective Confidence Scores (João Monteiro, Pau Rodriguez, Pierre-André Noël, Issam H. Laradji, David Vazquez)
Choreographer: Learning and Adapting Skills in Imagination (Pietro Mazzaglia, Tim Verbelen, Bart Dhoedt, Alexandre Lacoste, Sai Rajeswar Mudumba)
Breadth-First Pipeline Parallelism (Joel Lamy Poirier)
A General-Purpose Neural Architecture for Geospatial Systems (Alexandre Lacoste, Martin Weiss, Nasim Rahaman)
Attention for Compositional Modularity (Oleksiy Ostapenko, Alexandre Lacoste, Pau Rodriguez)
Countering Language Drift with KL Regularization. (Issam H. Laradji, Michael Noukhovitch, Aaron Courville)
Pierre-André Noël, senior applied research scientist, was one of the ServiceNow Research team members who presented their research poster at the ServiceNow booth.
David Vazquez, senior manager of research programs, welcomed attendees to the booth and answered their questions.
© 2022 ServiceNow, Inc. All rights reserved. ServiceNow, the ServiceNow logo, Now, and other ServiceNow marks are trademarks and/or registered trademarks of ServiceNow, Inc. in the United States and/or other countries. Other company names, product names, and logos may be trademarks of the respective companies with which they are associated.