Call for Papers: ICML 2020 Workshop on Uncertainty and Robustness in Deep Learning
Balaji Lakshminarayanan
ICML 2020 Workshop: Uncertainty and Robustness in Deep Learning
Virtual event (see here for more information)
July 17 or 18, 2020
https://sites.google.com/view/udlworkshop2020
DESCRIPTION:
There has been growing interest in rectifying deep neural network instabilities. Challenges arise when models receive samples drawn from outside the training distribution. For example, a neural network tasked with classifying handwritten digits may assign high confidence predictions to cat images. Anomalies are frequently encountered when deploying ML models in the real world. Well-calibrated predictive uncertainty estimates are indispensable for many machine learning applications, such as self-driving vehicles and medical diagnosis systems. Generalization to unseen and worst-case inputs is also essential for robustness to distributional shift. In order to have ML models reliably predict in open environment, we must deepen technical understanding in the emerging areas of: (1) learning algorithms that can detect changes in data distribution (e.g. out-of-distribution examples); (2) mechanisms to estimate and calibrate confidence produced by neural networks in typical and unforeseen scenarios; (3) methods to improve out-of-distribution generalization, including generalization to temporal, geographical, hardware, adversarial, and image-quality changes; (4) benchmark datasets and protocols for evaluating model performance under distribution shift; and (5) key applications of robust and uncertainty-aware deep learning (e.g., computer vision, robotics, self-driving vehicles, medical imaging) as well as broader machine learning tasks.
We invite the submission of papers on topics including, but not limited to:
Model uncertainty estimation and calibration
Probabilistic (Bayesian and non-Bayesian) neural networks
Anomaly detection and out-of-distribution detection
Robustness to distribution shift and out-of-distribution generalization
Model misspecification
Quantifying different types of uncertainty (known unknowns, unknown unknowns, contextual anomalies, ambiguities)
Open world recognition and open set learning
Connections between out-of-distribution generalization and adversarial robustness
New datasets and protocols for evaluating uncertainty and robustness
KEY DATES:
Submission deadline: May 22, 2020
Author notification: June 3, 2020
Camera ready deadline: July 1, 2020
Workshop date: July 17 or 18, 2020 (TBC)
SUBMISSION INSTRUCTIONS:
Researchers interested in contributing should upload a short paper of up to 4 pages in PDF format by May 22, 2020, 11:59pm (anywhere on earth) using https://easychair.org/conferences/?conf=udl2020. References and supplementary material can exceed 4 pages. Authors should use the style files in this zip file. Submissions don't need to be anonymized.
If you have any questions, please contact us at udlwork...@gmail.com.
ORGANIZERS:
Yixuan (Sharon) Li
Dan Hendrycks
Jasper Snoek
Thomas Dietterich
Balaji Lakshminarayanan