Lightnews — Scholar-powered news

Jakob Runge @jakobrunge.bsky.social · 7d

Save the date! 🌍

Climate Informatics will convene researchers at the intersection of weather & climate science, statistics, and machine learning, April 27-30, 2026 on our beautiful UNIL/EPFL campus by Lake Geneva in Lausanne, Switzerland.

More info: climateinformatics.org

Climate Informatics

An open community at the intersection of climate science and data science.

climateinformatics.org

Riccardo @rmassidda.it · Apr 2

🎉 The first edition of the "Causal Abstractions and Representations" workshop is here!

🇧🇷 Proudly hosted by @auai.org, we'll be in Rio de Janeiro on July 25, 2025.

🔨 Check out our invited speakers and the call for papers: we can't wait to see your submissions!

🌐 sites.google.com/view/car-25/

CAR

Causal Abstractions and RepresentationsWorkshop @ UAI 2025July 25th 2025, Rio de Janeiro 🇧🇷

sites.google.com

2 6

Jakob Runge @jakobrunge.bsky.social · Dec 17

Got multivariate X and/or Y and want to test conditional independence X _|_ Y | Z ? PairwiseMultCI is a wrapper that turns any univariate CI test into a multivariate one... it can also help increase power!
Tigramite tutorial github.com/jakobrunge/t... UAI Paper proceedings.mlr.press/v216/hochspr...

tigramite/tutorials/causal_discovery/tigramite_tutorial_pairwise_mult_ci.ipynb at master · jakobrunge/tigramite

Tigramite is a python package for causal inference with a focus on time series data. The Tigramite documentation is at - jakobrunge/tigramite

github.com

2 14

Jakob Runge @jakobrunge.bsky.social · Nov 29

Causal discovery loves conditional independence tests -- here's our CI test of the month: ParCorrWLS can deal with heteroskedastic data! Tigramite tutorial: github.com/jakobrunge/t... NeurIPS paper: proceedings.neurips.cc/paper_files/...

tigramite/tutorials/causal_discovery/tigramite_tutorial_heteroskedastic_ParCorrWLS.ipynb at master · jakobrunge/tigramite

Tigramite is a python package for causal inference with a focus on time series data. The Tigramite documentation is at - jakobrunge/tigramite

github.com

1 8

Jakob Runge @jakobrunge.bsky.social · Nov 21

Hello BlueSky! This is the start of my new timeline, gradually moving away from X/Twitter. Thanks for following! I will share new work here and look forward to exchanging ideas on causal inference for time series data with you!

18

Jakob Runge @jakobrunge.bsky.social · Nov 26

Bagged-PCMCI+ is out! A bootstrap approach to enhance precision+recall and confidence quantification for causal links. Check out our CLeaR paper: https://rb.gy/6t5gba and tigramite tutorial: https://rb.gy/db9ro1 #CLeaR2024 @ELLISforEurope @ERC_Research @KDebeire

1 4

Jakob Runge @jakobrunge.bsky.social · Nov 26

Contribute to the 2nd workshop on causal inference for time series at UAI, this year on July 19 in Barcelona! Deadline for submissions is May 19, less than 10 days to go! Looking forward to another successful event! Details: https://sites.google.com/view/ci4ts2024/home #CI4TS #UAI2024

Causal inference for time series

About the workshop

sites.google.com

Jakob Runge @jakobrunge.bsky.social · Nov 26

Got causal questions and multiple datasets collected for different contexts/conditions/subjects/locations? Try J-PCMCI+, a causal discovery method for multiple time-series datasets, able to handle both observed and latent context-confounding! https://proceedings.mlr.press/v216/gunther23a.html

1

Jakob Runge @jakobrunge.bsky.social · Nov 26

How do you choose you pet friend in the zoo of causal inference methods? Our recent JMLR paper provides a comprehensive benchmark comparison for the bivariate cause-effect challenge: https://www.jmlr.org/papers/v24/22-0151.html. For this and many more benchmarks visit http://causeme.net!

1 3

Jakob Runge @jakobrunge.bsky.social · Nov 26

A view-only version can be found here: https://rdcu.be/dfs5X

1 2

Jakob Runge @jakobrunge.bsky.social · Nov 26

3/3 Another highlight: A decision diagram to find the right approach for your problem. Great joint work with Andreas Gerhardus, Gherardo Varando, Veronika Eyring, and Gustau Camps-Valls. Integrate causal thinking into your #MachineLearning data-driven science!

Jakob Runge @jakobrunge.bsky.social · Nov 26

1/3 Just published @NatRevEarthEnv https://tinyurl.com/3zb8cu7s. A guide to #causalinference for time series: Phrase your problem as a causal Question, transparently state Assumptions, and apply the right method on your Data with the QAD-template based on @yudapearl's causal hierarchy

1 2 5

Jakob Runge @jakobrunge.bsky.social · Nov 26

Looking for a gentle introduction to #causalinference and relations to #MachineLearning learning? Kenneth Styppa heads a blog series of our group at https://medium.com/causality-in-data-science

Causality in Data Science – Medium

In this blog researchers and practitioners from the causal inference research group at the german aerospace center publish easy to read blog articles that should give an introduction to the topics of causal inference in machine learning.

medium.com

1 2

Jakob Runge @jakobrunge.bsky.social · Nov 26

We are hiring a scientific programmer (German E13 salary) in Jena to support tigramite and other software projects! Interested? More on https://climateinformaticslab.com/jobs/

Jobs

climateinformaticslab.com

Jakob Runge @jakobrunge.bsky.social · Nov 26

To all (too)late-submitters: We extended the paper deadline by a few days, it is now June 03 11:59AM UTC-0 !

Jakob Runge @jakobrunge.bsky.social · May 29

Few days left to submit to our UAI workshop on causal inference for time series data!

Jakob Runge @jakobrunge.bsky.social · Apr 27

Happy to announce a workshop on causal inference for time series data at @UncertaintyInAI in August this year, together with @saramagliacane @ckassaad @JonasChoice and others! Including a Call4Papers 👉 https://sites.google.com/view/ci4ts2023

Jakob Runge @jakobrunge.bsky.social · Nov 26

Few days left to submit to our UAI workshop on causal inference for time series data!

Jakob Runge @jakobrunge.bsky.social · Apr 27

Happy to announce a workshop on causal inference for time series data at @UncertaintyInAI in August this year, together with @saramagliacane @ckassaad @JonasChoice and others! Including a Call4Papers 👉 https://sites.google.com/view/ci4ts2023

Jakob Runge @jakobrunge.bsky.social · Nov 26

Happy to announce a workshop on causal inference for time series data at @UncertaintyInAI in August this year, together with @saramagliacane @ckassaad @JonasChoice and others! Including a Call4Papers 👉 https://sites.google.com/view/ci4ts2023

Jakob Runge @jakobrunge.bsky.social · Nov 26

Great to be attending @Climformatics in lovely Cambridge! Congratulations to the organisers for setting it up so well! https://twitter.com/Climformatics/status/1648607027671834624

Jakob Runge @jakobrunge.bsky.social · Nov 26

Join the Causal Inference Lab! Open #postdoc position on developing #causality methods for a range of application domains! --> http://climateinformaticslab.com @DLRdatascience @ELLISforEurope @ERC_Research

Jakob Runge @jakobrunge.bsky.social · Nov 26

2/2 The paper also introduces Mapped-PCMCI, a spatio-temporal causal discovery method to reconstruct causal networks from gridded data: Going from correlation to causal networks utilizing the assumption of a lower-dimensional latent causal process.

1

Jakob Runge @jakobrunge.bsky.social · Nov 26

1/2: We just published a paper in @EnvDataScience where we present the SAVAR model, a spatiotemporal toy model for benchmarking causal inference methods. Congratulations to Xavi Tibau! And now time to test your causal methods! https://bit.ly/3LQA7zh

1 1

Jakob Runge @jakobrunge.bsky.social · Nov 26

PS: We now have *flexible home office rules* at DLR enabling you to partially also work from another city than Jena.

Jakob Runge @jakobrunge.bsky.social · Nov 26

**Join the Causal Inference and Climate Informatics Lab!**
We are expanding with 4 open #postdoc positions (#phd also possible) on developing #causality theory and methods for #EarthSciences and beyond!
👉 http://climateinformaticslab.com
@DLRdatascience @ELLISforEurope @ERC_Research

News

climateinformaticslab.com

1

Jakob Runge @jakobrunge.bsky.social · Nov 26

2/2 Interested to develop #Causality and #AI theory and methods and work with climate scientists to better understand climate change and extremes? Join our team at http://climateinformaticslab.com

News

climateinformaticslab.com

Jakob Runge @jakobrunge.bsky.social · Nov 26

What are the causes behind recent extreme floods? Coming from math/stats/physics/ML and want to develop #Causality and #AI theory and methods to better understand extremes? --> Open positions with @ZscheischlerJak and further collaborators: http://climateinformaticslab.com @Compound_Event

1

Jakob Runge @jakobrunge.bsky.social · Nov 26

Got a background in math/stats/physics/ML and want to work on #Causality and #AI inspired by challenges in #EarthSciences and #ClimateChange? --> Well-funded #Postdoc/#PhD posts at @TUBerlin and @DLR_en Jena! --> http://climateinformaticslab.com @ELLISforEurope @ERC_Research

Jakob Runge @jakobrunge.bsky.social · Nov 26

The ClimateInformaticsLab starts a new branch at @TUBerlin as part of my ERC Starting Grant #CausalEarth. We have two upcoming #Postdoc/#PhD positions in #Berlin on #Causality and #AI for #Earth sciences. More info: https://climateinformaticslab.com @ELLISforEurope @ERC_Research @EU_H2020

News

climateinformaticslab.com

Jakob Runge @jakobrunge.bsky.social · Nov 26

3/3 LPCMCI is based on the flexible constraint-based causal discovery framework and can deal with linear and nonlinear, lagged and contemporaneous causal relationships. Now part of https://github.com/jakobrunge/tigramite

GitHub - jakobrunge/tigramite: Tigramite is a python package for causal inference with a focus on time series data. The Tigramite documentation is at

Tigramite is a python package for causal inference with a focus on time series data. The Tigramite documentation is at - jakobrunge/tigramite

github.com

Jakob Runge @jakobrunge.bsky.social · Nov 26

2/3 LPCMCI extends PCMCI+ (http://www.auai.org/uai2020/proceedings/579_main_paper.pdf) to the latent confounder case and shows strong gains in recall compared to previous methods for the ubiquitous high-autocorrelation case.

1

Jakob Runge @jakobrunge.bsky.social · Nov 26

Keep an eye on upcoming exciting PhD and postdoc opportunities within the ELLIS network that fosters AI in Europe. Soon I'll also announce some positions in my lab as part of my recent ERC Starting grant: https://ellis.eu/de/news/ellis-phd-program-call-for-applications