Finalizing the program for NeurIPS 2019
The NeurIPS conference is less than a month away! In fact, the pre-proceedings are now available online here, so people can get a head start at reading the papers of the conference.
We’ve also been working hard on putting the final touches to the conference program, and with this blog post would like to share what we’ve been up to.
This year, the NeurIPS conference will feature 4 parallel tracks of contributed presentations. With the growth in submissions, we determined that it was more appropriate to add an additional track to the three the conference had in 2018. The tracks will feature both oral (15 minutes) and spotlight (5 minutes) presentations, grouped together based on topic similarity. A survey was used to estimate the popularity of the various oral presentations, in order to determine which sessions should use the largest presentation rooms. But fear not, if a track overflows, it will be possible to watch the presentations online, either live through the livestream or afterwards on the recording.
In the poster sessions this year we’ve put in extra work in organizing the sessions thematically. We asked authors to identify a single (second-level) subject area for their work. Then, we used that information to create clusters of posters in each poster session. The most popular areas were split across multiple sessions, trying to avoid having more than 20 posters for a single area in any given poster session.
You will find the list of subject areas for each poster session on the NeurIPS website, here.
By clicking on a subject area, you’ll see which papers are being presented and the corresponding poster number. Similar information will also be made available at the conference, in various places. We hope this will make it easier for everyone to find the content that is most relevant to their research interests.
Online discussion of papers on OpenReview
To further facilitate interactions between authors of contributed papers and the rest of the NeurIPS community, we are also encouraging researchers to discuss papers on the following OpenReview site:
The site was put up by the Reproducibility Chairs (Joelle Pineau and Koustuv Sinha), as part of the Reproducibility Challenge they are running. However, we encourage everyone to also use it as a platform for general discussion and to post questions you may have about any contributed paper.
For papers featured as oral presentations at the conference, we will encourage the session chairs (who will moderate questions after oral presentations) to pick one question posted on OpenReview to ask the speaker. If you’d like your question to be considered, please visit the OpenReview page for the corresponding paper and post it there. Here’s the list of orals, with a link to its page:
- Uniform convergence may be unable to explain generalization in deep learning
- Logarithmic Regret for Online Control
- Kernel Instrumental Variable Regression
- Updates of Equilibrium Prop Match Gradients of Backprop Through Time in an RNN with Static Input
- Generative Modeling by Estimating Gradients of the Data Distribution
- HYPE: A Benchmark for Human eYe Perceptual Evaluation of Generative Models
- Trajectory of Alternating Direction Method of Multipliers and Adaptive Acceleration
- Necessary and Sufficient Geometries for Gradient Methods
- Causal Confusion in Imitation Learning
- Using a Logarithmic Mapping to Enable Lower Discount Factors in Reinforcement Learning
- Scalable Bayesian inference of dendritic voltage via spatiotemporal recurrent state space models
- Parameter elimination in particle Gibbs sampling
- Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representations
- A neurally plausible model learns successor representations in partially observable environments
- Fast and Accurate Least-Mean-Squares Solvers
- Faster width-dependent algorithm for mixed packing and covering LPs
- Dynamics of stochastic gradient descent for two-layer neural networks in the teacher-student setup
- On Making Stochastic Classifiers Deterministic
- Batched Multi-armed Bandits Problem
- Strategizing against No-regret Learners
- Guided Similarity Separation for Image Retrieval
- Geometry-Aware Neural Rendering
- Efficient and Thrifty Voting by Any Means Necessary
- Distribution-Independent PAC Learning of Halfspaces with Massart Noise
- Brain-Like Object Recognition with High-Performing Shallow Recurrent ANNs
- Variance Reduction for Matrix Games
- Exponentially convergent stochastic k-PCA without variance reduction
- Optimizing Generalized Rate Metrics with Three Players
- Average Individual Fairness: Algorithms, Generalization and Experiments
- On Robustness of Principal Component Regression
- Putting An End to End-to-End: Gradient-Isolated Learning of Representations
- XLNet: Generalized Autoregressive Pretraining for Language Understanding
- Blind Super-Resolution Kernel Estimation using an Internal-GAN
- Nonparametric Density Estimation & Convergence Rates for GANs under Besov IPM Losses
- Understanding Sparse JL for Feature Hashing
- R2D2: Reliable and Repeatable Detector and Descriptor
In addition to contributed presentations, NeurIPS also features keynote presentations by recognized experts from various domains relevant to our community. Among these presentations is the Posner Lecture (in honor of Ed Posner, the first president of the NeurIPS Foundation). The Posner Lecture is usually reserved to a long-time contributor to the NeurIPS conference. The Breiman Lecture (in honor of statistician Leo Breiman, who served on the NeurIPS Board for more than 10 years) is dedicated to work in statistics relevant to the NeurIPS community. It’s common for the other NeurIPS keynotes to span a broad spectrum of topics, such as neuroscience, cognitive science, biology, data science, and many more.
This year, we’re excited to have the following 7 speakers:
- Blaise Aguera y Arcas
- Yoshua Bengio (Posner Lecture)
- Kafui Dzirasa
- Celeste Kidd
- Jeff Heer
- Dana Pe’er
- Bin Yu (Breiman Lecture)
The final selection was made after coordination with the Diversity and Inclusion Chairs (Katherine Heller and Charles Isbell) and the General Chair (Hanna Wallach), so that as many dimensions of diversity as possible were taken into account. See this page for more information on the title and description of each keynote.
The last bit of information we’ll want to share with you about the program are the paper awards. We will unveil these closer to the conference, so stay tuned for that!
Alina Beygelzimer, Emily Fox, Florence d’Alché-Buc, Hugo Larochelle
NeurIPS 2019 Program Chairs