What You Need to Know About NIPS 2017

The Conference on Neural Information Processing (NIPS) kicks off today in Long Beach, California and runs to December 9.

The NIPS 2017 speaker list is a who’s who of AI: Yann LeCun from Facebook, Demis Hassabis and David Silver from DeepMind, Gary Marcus from New York University, Juergan Shimidhubor from Switzerland-based Dalle Molle Institute for Artificial Intelligence Research, etc.

NIPS 2017 is entirely sold out and did so with unprecedented speed. Facebook AI researcher Alex LeBurn posted a graph on Twitter illustrating the distribution trend of NIPS conference registrations from 2002 to 2017.

First held in 1987, NIPS was initially an interdisciplinary meeting for researchers exploring biological and artificial neural networks. Since the biological and artificial systems research streams diverged in the 1990s, NIPS has focused on machine learning, AI and statistics.

This is the first time since 2013 that NIPS has been held in the US, and may be the last for awhile given travel issues in that country. From the NIPS website: “Our goal is to open NIPS Conferences and Workshops to researchers from all parts of the world and we regret that US government restrictions may make it difficult or impossible for some to participate this year.” NIPS was held in Barcelona last year and it will move to Montreal in 2018.

A record-high 3,240 papers were submitted to NIPS this year, with 678 selected. The 2017 NIPS Best Paper Award goes to CMU Professor Tuomas Sandholm and his student Noam Brown for their paper demonstrating how an AI-powered poker computer (Libratus) beat top human players.

NIPS 2017 has five competition tracks: Learning to Run, Deep RL; Human-Computer Question Answering; Classifying Clinically Actionable Genetic Mutations; Adversarial Attacks and Defences; and the Conversational Intelligence Challenge.

Synced will be reporting from NIPS throughout the week.

Journalist: Tony Peng | Editor: Michael Sarazen