NIPS 2019 is approaching. In order to make the best use of this jam-packed 7-day conference, I made a rough plan that I will lay out in the following sections. My motivation for NIPS 2019 is influenced by my interests in non-deterministic environments, encoding complexity in reinforcement learning, and techniques that (might) lead to human-level artificial intelligence. This schedule is biased towards self-supervised learning (SSL) and some reinforcement learning (RL).
Day 1: NIPS 2019 Expo (Dec 8)
NIPS 2019 has a great Expo program that brings together industry players to share how they are contributing to research and implementations. There are 14 consecutive Talks and Panels, two sessions of Workshops with four and five workshops in parallel, and nine Demos from industry partners.
With so much going on in just a single day, you really have to manage your time carefully. I will try to attend Interpretability — now what? and Explainable Reinforcement Learning as well as the following workshops: Causal Inference & Reinforcement learning: Making the right intervention, Responsible and Reproducible AI with PyTorch and Facebook, and Real world reinforcement learning with Vowpal Wabbit.
Day 2: Tutorials (Dec 9)
Although all NIPS tutorials this year sound excellent. I will focus my attention on the following three: Imitation Learning and its Application to Natural Language Generation, Human Behavior Modeling with Machine Learning: Opportunities and Challenges, and Reinforcement Learning: Past, Present, and Future Perspectives.
Day 3–5: Invited Talks, Papers, and Demos (Dec 10–12)
I am very excited about two invited talks at this year’s NIPS:
- Celeste Kidd talks about core cognitive systems humans use and what it tells us about how to design artificial systems in How to Know;
- Yoshua Bengio talks about the transition to building systems for conscious task humans perform in From System 1 Deep Learning to System 2 Deep Learning.
I have picked a couple of papers that I am reading before NIPS:
- Andre Barreto, Diana Borsa, Shaobo Hou, Gheorghe Comanici, Eser Aygün, Philippe Hamel, Daniel Toyama, Jonathan hunt, Shibl Mourad, David Silver, and Doina Precup. The Option Keyboard: Combining Skills in Reinforcement Learning.
- John Lawson, George Tucker, Bo Dai, and Rajesh Ranganath. Energy-Inspired Models: Learning with Sampler-Induced Distributions.
- Simon Ramstedt and Chris Pal. Real-Time Reinforcement Learning.
- Shikun Liu, Andrew Davison, and Edward Johns. Self-Supervised Generalisation with Meta Auxiliary Learning.
- Patrick Putzky and Max Welling. Invert to Learn to Invert.
- Xueting Li, Sifei Liu, Shalini De Mello, Xiaolong Wang, Jan Kautz, and Ming-Hsuan Yang. Joint-task Self-supervised Learning for Temporal Correspondence.
- Tam Nguyen, Maximilian Dax, Chaithanya Kumar Mummadi, Nhung Ngo, Thi Hoai Phuong Nguyen, Zhongyu Lou, and Thomas Brox. DeepUSPS: Deep Robust Unsupervised Saliency Prediction via Self-supervision.
Day 6–7: Workshops (Dec 13–14)
There are workshops for everyone at NIPS. The ones I will focus on are:
- KR2ML — Knowledge Representation and Reasoning Meets Machine Learning
- Bayesian Deep Learning
- Solving inverse problems with deep networks: New architectures, theoretical foundations, and applications
- Biological and Artificial Reinforcement Learning
- Graph Representation Learning
- “Do the right thing”: machine learning and causal inference for improved decision making
- Deep Reinforcement Learning
Misc: Socials and Meetups
There are 15 socials from Tuesday to Thursday.
UPDATE: I was so excited for NIPS but the regular “mobile-friendly” schedule was a bit overwhelming, so I created a Airtable view with the entire program (including speakers and descriptions) so I can customize my schedule for NIPS 2019: https://airtable.com/universe/expwBXEowz7SW9P3E/nips-2019-conference-schedule
Which side of the “cake battle” are you on?
I wish everyone an exciting NIPS 2019.