![[2507.12257] Robust Causal Discovery in Real-World Time Series with Power-Laws](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2507.12257] Robust Causal Discovery in Real-World Time Series with Power-Laws
Summary
This paper presents a novel method for causal discovery in time series data, leveraging power-law distributions to enhance robustness against noise and improve causal inference accuracy.
Why It Matters
Understanding causal relationships in time series is crucial across various fields such as finance and neuroscience. This research addresses the limitations of existing causal discovery algorithms, offering a more reliable approach that can lead to better decision-making and insights in real-world applications.
Key Takeaways
- The proposed method utilizes power-law spectral features to improve causal discovery.
- It outperforms existing algorithms on both synthetic and real-world datasets.
- The research highlights the importance of robustness in causal inference.
- Applications span multiple domains including finance and climate science.
- The findings could lead to more accurate modeling of complex systems.
Computer Science > Machine Learning arXiv:2507.12257 (cs) [Submitted on 16 Jul 2025 (v1), last revised 17 Feb 2026 (this version, v3)] Title:Robust Causal Discovery in Real-World Time Series with Power-Laws Authors:Matteo Tusoni, Giuseppe Masi, Andrea Coletta, Aldo Glielmo, Viviana Arrigoni, Novella Bartolini View a PDF of the paper titled Robust Causal Discovery in Real-World Time Series with Power-Laws, by Matteo Tusoni and 5 other authors View PDF HTML (experimental) Abstract:Exploring causal relationships in stochastic time series is a challenging yet crucial task with a vast range of applications, including finance, economics, neuroscience, and climate science. Many algorithms for Causal Discovery (CD) have been proposed; however, they often exhibit a high sensitivity to noise, resulting in spurious causal inferences in real data. In this paper, we observe that the frequency spectra of many real-world time series follow a power-law distribution, notably due to an inherent self-organizing behavior. Leveraging this insight, we build a robust CD method based on the extraction of power-law spectral features that amplify genuine causal signals. Our method consistently outperforms state-of-the-art alternatives on both synthetic benchmarks and real-world datasets with known causal structures, demonstrating its robustness and practical relevance. Subjects: Machine Learning (cs.LG); Data Analysis, Statistics and Probability (physics.data-an); Machine Learning (stat.ML); Other ...
