Simulation-based Inference for scientific discovery

Europe/Berlin
Online via Zoom

Online via Zoom

Description

You use simulation in physics, economics, archaeology or any other domain of your choice? You want to find the simulator's parameters that best fits the observations? Then simulation-based inference is something for you!

We teach a practical simulation-based inference workshop to help you understand recent machine learning techniques and apply them to your problem. We plan to cover the following topics:

  • Day 1: Using simulators for discovery, introduction to conditional density estimation.
    • a discussion of simulators
    • parameter inference
    • Bayesian Inference without Likelihoods
  • Day 2: How does neural simulation-based inference work?
    • Bayes' Rule with neural density estimators
    • posterior estimation (SNPE)
    • likelihood estimation (SNLE and beyond)
  • Day 3: Applying the sbi toolbox to your problem. Pitfalls, tricks and opportunities!
    • hands-on: sbi
    • hands-on: sbi and normalizing flows
    • hands-on: calibration or misspecification
    • bring your own data

Apply to learn, have fun, and participate in a supportive and inclusive community. The workshop will combine lectures and practical hands-on sessions by experts in the field. We strive to provide a seamless computing environment for you to focus on the content rather than in import errors.

Code of Conduct

We believe that a diverse and welcoming environment is key for a successful workshop, regardless of background or identity. To ensure that everyone in our event feel comfortable, supported and respected, we elaborated a Code of Conduct, you accept to abide by while taking part in this workshop.

Interested?

Due to teaching capacity, we can only select 20 participants for the workshop. To come up with that list, we have formed a small selection board and made this decision until end of August, 2021. We will inform successful applicants subsequently.

When & Where?

20-22 September, 9am - 5pm CEST

We will circulate the zoom link to all accepted learners.

Any other Questions?

Join our simulation-based inference discussion forum in Zulip (you are welcome whether a participant or not!)
    

Instructors and Support

Jan-Matthis Lückmann

Jan-Matthis Lückmann
mackelab @ Tübingen University

Daniela Huppenkothen

Daniela Huppenkothen
SRON Netherlands Institute for Space Research

David Greenberg

David Greenberg
Helmholtz AI & Helmholtz-Zentrum Hereon

Jan Bölts

Jan Bölts
mackelab @ Tübingen University

Michael Deistler

Michael Deistler
mackelab @ Tübingen University

Peter Steinbach

Peter Steinbach
Helmholtz AI & Helmholtz-Zentrum Dresden-Rossendorf

Álvaro Tejero-Cantero

Álvaro Tejero-Cantero
mlcolab @ Tübingen University

Pedro J. Gonçalves

Pedro J. Gonçalves
mackelab @ caesar

 

Organisational Team

Álvaro Tejero-Cantero

Álvaro Tejero-Cantero
mlcolab @ Tübingen University

Elena Sizana

Elena Sizana
mlcolab @ Tübingen University

Stefan Wezel

Stefan Wezel
mlcolab @ Tübingen University

Peter Steinbach

Peter Steinbach
Helmholtz AI & Helmholtz-Zentrum Dresden-Rossendorf

Supporting Organisations

Helmholtz AI

mlcolab

mackelab

mlcluster

Uni Tuebingen

    • 1
      Welcome
      Speakers: Mr Alvaro Tejero-Cantero (MLcolab), Peter Steinbach (HZDR)
    • 2
      Simulators for Science
      Speaker: Mr Alvaro Tejero-Cantero (MLcolab)
    • 3
      Introduce your simulators
      Speakers: Mr Alvaro Tejero-Cantero (MLcolab), Peter Steinbach (HZDR)
    • 4
      From Simulators to Simulations-based Inference
      Speaker: Mr Alvaro Tejero-Cantero (MLcolab)
    • 12:00 PM
      Lunch
    • 5
      Computing on HAICORE
      Speaker: Stefan Kesselheim (FZ Jülich)
    • 6
      Practical: ABC hands-on
      Speakers: Mr Alvaro Tejero-Cantero (MLcolab), Stefan Kesselheim (FZ Jülich)
    • 7
      Lecture: SBI in neuroscience
      Speaker: Mr Pedro Goncalves (Mackelab)
    • 8
      Learner Feedback Day 1
      Speaker: Peter Steinbach (HZDR)
    • 9
      Conditional Density Estimation
      Speaker: Michael Deistler (MackeLab)
    • 10
      Introduction to SNPE
      Speakers: Mr David Greenberg (Hereon), Michael Deistler (MackeLab)
    • 12:00 PM
      Lunch
    • 11
      SNLE and SNRE
      Speaker: Mr David Greenberg (Hereon)
    • 12
      Practical: The SBI package
      Speaker: Mr Jan Bölts
    • 13
      Learner Feedback Day 2
    • 14
      Benchmarking SBI
      Speaker: Mr Jan-Matthis Lückmann (MackeLab)
    • 15
      Troubleshooting SBI
      Speaker: Mr Jan Bölts
    • 16
      Bayesian Workflow
      Speaker: Mr Jan Bölts
    • 12:00 PM
      Lunch
    • 17
      SBI with your data
      Speakers: Mr Alvaro Tejero-Cantero (MLcolab), Mr Jan Bölts, Mr Jan-Matthis Lückmann (MackeLab), Michael Deistler (MackeLab), Peter Steinbach (HZDR)