AGU Session on self-supervised methods in the Earth Sciences (GC099)
17 views
Skip to first unread message
Elisabeth Moyer
Jul 14, 2023, 2:33:14 PM7/14/23
to Climate Informatics News
Dear colleagues -
I want to call your attention to a session on self-supervised methods in the Earth Sciences, an area of great promise across many fields. The session description is below, and we are delighted to welcome Claire Monteleoni and Pierre Gentine as invited speakers. Because work in this area is relatively new, we welcome abstracts that describe preliminary or prospective work as well as mature research. We hope this session will serve to connect researchers across many areas to build the research community. For that reason, we also hope to cast a wide net in advertising the session: please do forward to colleagues who might be interested.
best,
Liz Moyer (with Ian Foster, Jim Franke, and Takuya Kurihana)
GC099. Self-Supervised Deep Learning Applications in the Earth Sciences
The dramatic growth in data volume from both numerical simulations and satellite observations has prompted interest in using deep learning applications to provide some form of automated analysis or categorization of large datasets. However, for most Earth science datasets, pre-existing labels or targets are generally not plausible, or even desirable. The greater need is to learn relevant categories or distinctions without explicit supervision. Unsupervised learning approaches have the potential to unlock the power of large datasets in areas such as cloud physics, geomorphology, ecology, and even paleontology. These approaches do have challenges, including validating the utility of categories produced by unsupervised classification, or extending the approach to fit meaningful generative models. This session welcomes submissions on new methodologies, challenges, and scientific results using self-supervised or unsupervised deep learning applications, in climate in particular but across all of Earth sciences.
For some recent work on applications of self-supervised learning to cloud classification, see:
https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9497325
https://www.mdpi.com/2072-4292/14/22/5690/htm
I want to call your attention to a session on self-supervised methods in the Earth Sciences, an area of great promise across many fields. The session description is below, and we are delighted to welcome Claire Monteleoni and Pierre Gentine as invited speakers. Because work in this area is relatively new, we welcome abstracts that describe preliminary or prospective work as well as mature research. We hope this session will serve to connect researchers across many areas to build the research community. For that reason, we also hope to cast a wide net in advertising the session: please do forward to colleagues who might be interested.
best,
Liz Moyer (with Ian Foster, Jim Franke, and Takuya Kurihana)
GC099. Self-Supervised Deep Learning Applications in the Earth Sciences
The dramatic growth in data volume from both numerical simulations and satellite observations has prompted interest in using deep learning applications to provide some form of automated analysis or categorization of large datasets. However, for most Earth science datasets, pre-existing labels or targets are generally not plausible, or even desirable. The greater need is to learn relevant categories or distinctions without explicit supervision. Unsupervised learning approaches have the potential to unlock the power of large datasets in areas such as cloud physics, geomorphology, ecology, and even paleontology. These approaches do have challenges, including validating the utility of categories produced by unsupervised classification, or extending the approach to fit meaningful generative models. This session welcomes submissions on new methodologies, challenges, and scientific results using self-supervised or unsupervised deep learning applications, in climate in particular but across all of Earth sciences.
For some recent work on applications of self-supervised learning to cloud classification, see:
https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9497325
https://www.mdpi.com/2072-4292/14/22/5690/htm
Reply all
Reply to author
Forward
0 new messages