Program on Mathematical and Statistical Methods for Climate and the Earth System Opening Workshop, Discussion: Session on Remote Sensing - Amy Braverman, Aug 22, 2017
1 like272 views
The document discusses a session on remote sensing held at JPL, focusing on the role of statistical methods in handling massive data sets. Key presentations addressed the hierarchy of scientific questions, processing infrastructure constraints, and approaches to managing costs versus inferential quality. The working group's goal is to develop methods applicable within large-scale data processing frameworks, emphasizing uncertainty-cost trade-offs.
Related topics:1 of 7
Download to read offline
More Related Content
What's hot (20)
Ad
Viewers also liked (20)
Ad
Similar to Program on Mathematical and Statistical Methods for Climate and the Earth System Opening Workshop, Discussion: Session on Remote Sensing - Amy Braverman, Aug 22, 2017 (20)
More from The Statistical and Applied Mathematical Sciences Institute (20)
Recently uploaded (20)
Program on Mathematical and Statistical Methods for Climate and the Earth System Opening Workshop, Discussion: Session on Remote Sensing - Amy Braverman, Aug 22, 2017
- 1. Discussion: Session on Remote Sensing Amy Braverman Jet Propulsion Laboratory, California Institute of Technology August 22, 2017 1
- 2. Noel’s talk: broad overview of the role of statistical principles in remote sensing Dan’s talk: NASA’s remote sensing data processing and distribution infrastructure Matthias’ talk: framework for spatial statistics suitable for massive and distributed data 2
- 3. Noel’s: hierarchy of science questions paralleled by hierarchy of processing Dan’s talk: hierarchy of constraints imposed by processing infrastructure Matthias’ talk: hierarchy of solutions? 3
- 4. How do we connect these concepts in a way that is specific enough to suggest tangible research to pursue? Big constraint: can’t move (all) data Points to a need to flexibly trade-off inferential quality against cost (computational and transportation, etc.) 4
- 5. Matthias: example of new statistical methods that explicitly recognize the need to trade-off cost vs. inferential quality. General topic for the Working Group: develop/study methods that can be implemented within a large-scale data processing framework Requires understanding uncertainty-cost trade space 5
- 6. Some type of problems: 1. Estimating and modelling distributions in space and time with massive observational data volumes, and “massively complex" structure 2. Get away from Gaussians 3. Approximate sufficient statistics? Work by Petruta Caragea and Richard Smith 4. Faster MCMC and quasi Monte Carlo methods; multi-level Monte Carlo (Giles, 2015) 6
- 7. Many thanks for JPL’s Center for Data Science and Technology and Caltech’s Center for Data Driven Discovery for supporting these activities. ©2017 California Institute of Technology. Government sponsorship acknowledged. 7