CausalRivers: Causal Discovery Benchmark Poster
A conference poster introducing CausalRivers, a benchmark for causal discovery on real-world time-series data from river discharge stations in Germany, featuring ground truth graphs, distributional shifts, and benchmarking results of various methods.
Paper context
Paper title: CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series Abstract: A conference poster introducing CausalRivers, a benchmark for causal discovery on real-world time-series data from river discharge stations in Germany, featuring ground truth graphs, distributional shifts, and benchmarking results of various methods. Paper body (method & results): <!DOCTYPE html> <html lang="en"> <head> <meta content="text/html; charset=utf-8" http-equiv="content-type"/> <title>CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series</title> <!--Generated on Fri Mar 21 17:55:47 2025 by LaTeXML (version 0.8.8) http://dlmf.nist.gov/LaTeXML/.--> <meta content="width=device-width, initial-scale=1, shrink-to-fit=no" name="viewport"/> <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/css/bootstrap.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/ar5iv-fonts.0.7.9.min.css" rel="stylesheet" type="text/css"/> <link href="/static/browse/0.3.4/css/latexml_styles.css" rel="stylesheet" type="text/css"/> <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.0/dist/js/bootstrap.bundle.min.js"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/html2canvas/1.3.3/html2canvas.min.js"></script> <script src="/static/browse/0.3.4/js/addons_new.js"></script> <script src="/static/browse/0.3.4/js/feedbackOverlay.js"></script> <base href="/html/2503.17452v1/"/></head> <body> <nav class="ltx_page_navbar"> <nav class="ltx_TOC"> <ol class="ltx_toclist"> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S1" title="In CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">1 </span>Introduction</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S2" title="In CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">2 </span>Background</span></a></li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S3" title="In CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3 </span>Benchmark description</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S3.SS1" title="In 3 Benchmark description ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.1 </span>Benchmark construction</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S3.SS2" title="In 3 Benchmark description ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.2 </span>Benchmarking kit features</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S3.SS3" title="In 3 Benchmark description ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.3 </span>Baseline strategies</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S3.SS4" title="In 3 Benchmark description ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">3.4 </span>Unique Characteristics</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"> <a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S4" title="In CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4 </span>Experimental Results and Discussion</span></a> <ol class="ltx_toclist ltx_toclist_section"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S4.SS1" title="In 4 Experimental Results and Discussion ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.1 </span>Experiment Set 1 - Varying graph structures</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S4.SS2" title="In 4 Experimental Results and Discussion ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.2 </span>Experiment Set 2 - time-series subsampling</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S4.SS3" title="In 4 Experimental Results and Discussion ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">4.3 </span>Experiment Set 3 - Domain adaption</span></a></li> </ol> </li> <li class="ltx_tocentry ltx_tocentry_section"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#S5" title="In CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">5 </span>Conclusion</span></a></li> <li class="ltx_tocentry ltx_tocentry_appendix"> <a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#A1" title="In CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">A </span>Appendix</span></a> <ol class="ltx_toclist ltx_toclist_appendix"> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#A1.SS1" title="In Appendix A Appendix ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">A.1 </span>Data origins, preprocessing and data statistics</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#A1.SS2" title="In Appendix A Appendix ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">A.2 </span>Hyperparameters</span></a></li> <li class="ltx_tocentry ltx_tocentry_subsection"><a class="ltx_ref" href="https://arxiv.org/html/2503.17452v1#A1.SS3" title="In Appendix A Appendix ‣ CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series"><span class="ltx_text ltx_ref_title"><span class="ltx_tag ltx_tag_ref">A.3 </span>Additional resources - Experiment Set 1 - 3</span></a></li> </ol> </li> </ol></nav> </nav> <div class="ltx_page_main"> <div class="ltx_page_content"> <article class="ltx_document ltx_authors_1line"> <h1 class="ltx_title ltx_title_document">CausalRivers - Scaling up benchmarking of causal discovery for real-world time-series</h1> <div class="ltx_authors"> <span class="ltx_creator ltx_role_author"> <span class="ltx_personname">Gideon Stein, Maha Shadaydeh, Jan Blunk, Niklas Penzel, Joachim Denzler <br class="ltx_break"/>Computer Vision Group Jena <br class="ltx_break"/>Friedrich Schiller University Jena <br class="ltx_break"/>Jena, Thuringia 07743, Germany <br class="ltx_break"/><span class="ltx_text ltx_font_typewriter" id="id1.1.id1">gideon.stein@uni-jena.de</span> <br class="ltx_break"/> </span></span> </div> <div class="ltx_abstract"> <h6 class="ltx_title ltx_title_abstract">Abstract</h6> <p class="ltx_p" id="id2.id1">Causal discovery, or identifying causal relationships from observational data, is a notoriously challenging task, with numerous methods proposed to tackle it. Despite this, in-the-wild evaluation of these methods is still lacking, as works frequently rely on synthetic data evaluation and sparse real-world examples under critical theoretical assumptions. Real-world causal structures, however, are often complex, evolving over time, non-linear, and influenced by unobserved factors, making it hard to decide on a proper causal discovery strategy. To bridge this gap, we introduce <span class="ltx_text ltx_font_bold" id="id2.id1.1">CausalRivers<span class="ltx_note ltx_role_footnote" id="footnote1"><sup class="ltx_note_mark">1</sup><span class="ltx_note_outer"><span class="ltx_note_content"><sup class="ltx_note_mark">1</sup><span class="ltx_tag ltx_tag_note"><span class="ltx_text ltx_font_medium" id="footnote1.1.1.1">1</span></span><a class="ltx_ref ltx_href ltx_font_medium" href="https://causalrivers.github.io" title="">https://causalrivers.github.io</a></span></span></span></span>, the largest in-the-wild causal discovery benchmarking kit for time-series data to date. CausalRivers features an extensive dataset on river discharge that covers the eastern German territory (666 measurement stations) and the state of Bavaria (494 measurement stations). It spans the years 2019 to 2023 with a 15-minute temporal resolution. Further, we provide additional data from a flood around the Elbe River, as an event with a pronounced distributional shift. Leveraging multiple sources of information and time-series meta-data, we constructed two distinct causal ground truth graphs (Bavaria and eastern Germany). These graphs can be sampled to generate thousands of subgraphs to benchmark causal discovery across diverse and challenging settings. To demonstrate the utility of CausalRivers, we evaluate several causal discovery approaches through a set of experiments to identify areas for improvement. CausalRivers has the potential to facilitate robust evaluations and comparisons of causal discovery methods. Besides this primary purpose, we also expect that this dataset will be relevant for connected areas of research, such as time-series forecasting and anomaly detection. Based on this, we hope to push benchmark-driven method development that fosters advanced techniques for causal discovery, as is the case for many other areas of machine learning.</p> </div> <section class="ltx_section" id="S1"> <h2 class="ltx_title ltx_title_section"> <span class="ltx_tag ltx_tag_section">1 </span>Introduction</h2> <figure class="ltx_figure" id="S1.F1"> <div class="ltx_flex_figure"> <div class="ltx_flex_cell ltx_flex_size_1"> <figure class="ltx_figure ltx_figure_panel ltx_align_center" id="S1.F1.sf1"><img alt="Refer to caption" class="ltx_graphics ltx_img_portrait" height="1155" id="S1.F1.sf1.g1" src="x1.png" width="822"/> <figcaption class="ltx_caption"><span class="ltx_tag ltx_tag_figure">(a) </span>Eastern Germany</figcaption> </figur
