Bayes for scattering workshop

Europe/Zurich
M5 (COBIS)

M5

COBIS

Ole maaløes vej 3, Copenhagen
Registration
Participants
    • 12:00 PM 12:45 PM
      Lunch 45m
    • 12:45 PM 2:20 PM
      Use cases for Bayes in scattering
      • 12:45 PM
        Introduction 15m

        Why are we having this meeting?
        Brief intro to Bayes' theorem.

      • 1:00 PM
        Analysis of Small-Angle Scattering Data Using Model Fitting and Bayesian Regularization 20m
        Speaker: Andreas Larsen (University of Copenhagen)
      • 1:20 PM
        A simple bayesian method for 2D SANS 20m
        Speaker: Alexander Holmes (European Spallation Source ERIC)
      • 1:40 PM
        Bayes for reflectometry 20m
        Speaker: Andrew McCluskey (Diamond Light Source)
      • 2:00 PM
        Bayesian statistic for SAS and beyond 10m

        As a primer for discussion we will try to cover a few more SAS cases as well as other scattering techniques (particularly spectroscopy)

        Speaker: Wojciech Potrzebowski (European Spallation Source ERIC)
    • 2:00 PM 4:00 PM
      Discussion
      • 2:10 PM
        Break 5m
      • 2:15 PM
        Introduction 10m

        We will try out OST as at the NSS retreat
        https://confluence.esss.lu.se/display/SD/Science+Retreat+2018+-+Fundamentals+of+OST
        Think of topics before.
        Ideas
        Use cases
        Workflow of data analysis
        Presenting results
        Computational techniques

      • 2:25 PM
        Group discussions 35m
      • 3:00 PM
        Coffee break 10m
      • 3:10 PM
        Group discussions 50m
    • 4:00 PM 5:00 PM
      Wrap/up: Wrap-up
      • 4:00 PM
        Report from discussions 30m

        Items discussed at the meeting:

        1. Bayesian statistics to optimize experiments i.e. when do I have good statistics to stop measuring?
        2. How to treat variables correlations
        3. Resolution functions for small angle scattering
        4. Bayesian model comparison (how to select good model for fitting)
        5. Bayesian statistics as way to avoid overfitting
        6. Bayesian statistics and its relation to machine and deep learning
        7. How to make scattering communities aware of Bayesian statistics: constant presentation, highlight important results, show probability distribution
        8. Plotting distributions, Library for plotting
        9. Easy accessible software
        10. Can we have a "community prior"?
        11. Data format for prior and posterior to update
      • 4:30 PM
        Next steps 30m