Accepted papers
Oral presentation
Authors: Goutham Rajendran, Patrik Reizinger, Wieland Brendel, Pradeep Kumar Ravikumar
Title: An Interventional Perspective on Identifiability in Gaussian LTI Systems with Independent Component Analysis
Abstract: This study investigates the relationship between system identification and intervention design in dynamical systems. While previous research has demonstrated how identifiable representation learning methods, such as Independent Component Analysis (ICA), can reveal cause-effect relationships, it relied on a passive perspective without considering how to collect data. In this work, we demonstrate that in Gaussian Linear Time-Invariant (LTI) systems, the system parameters can be identified by introducing diverse intervention signals in a multi-environment setting. By harnessing appropriate diversity assumptions motivated by the ICA literature, our findings connect experiment design and representational identifiability in dynamical systems.
identifiability_data_rajendran.pdf2_GouthamRajendran.pdfOral presentation
Authors: Zsigmond Benkő, Marcell Stippinger, Attila Bencze, András Telcs, Zoltán Somogyvári
Title: Reconstruction of hidden common driver dynamics by anisotropic self-organizing neural networks
Abstract: We are introducing a novel approach to accurately reconstruct the underlying dynamics of hidden common drivers, based on analyzing time series data from the driven dynamical systems. The reconstruction process relies on time-delay embedding, estimation of the intrinsic dimension of the observed systems, and their mutual dimension. A key component of our approach is a new anisotropic training technique applied to Kohonen's self-organizing map, which effectively learns the attractor of the driven system and separates it into self-dynamics and shared dynamics.
6_reconstruction_of_hidden_commo.pdf4_Somogyvári23CI4TS.pdfOral presentation
Authors: Shanyun Gao, Raghavendra Addanki, Tong Yu, Ryan A. Rossi, Murat Kocaoglu
Title: Causal Discovery in Semi-Stationary Time Series
Abstract: Discovering causal relations from observational time series without making the stationary assumption is a significant challenge. In practice, this challenge is common in many areas, such as retail sales, transportation systems, and medical science. Here, we consider this problem for a class of non-stationary time series problems. The structural causal model (SCM) of this type of time series, called the semi-stationary time series, exhibits that a finite number of different causal mechanisms occur sequentially and periodically across time. This model holds considerable practical utility because it can represent periodicity, including common occurrences such as seasonality and diurnal variation. We propose a constraint-based, non-parametric algorithm for discovering causal relations in this setting. The resulting algorithm, PCMCIΩ, can capture the alternating and recurring changes in the causal mechanisms and then identify the underlying causal graph with conditional independence (CI) tests. We show that this algorithm is sound in identifying causal relations on discrete time series. We validate the algorithm with extensive experiments on continuous and discrete simulated data. We also apply our algorithm to a real-world climate dataset.
5_causal_discovery_in_semi_stati.pdf6_UAI workshop Shanyun Gao.pdfPoster
Authors: Yair Daon, Amit Huppert, Uri Obolski
Title: Refuting False Causal Relations in Epidemiological Time Series
Abstract: Causal detection is a crucial problem in epidemiology, where determining causes to outbreaks, immunity and chronic diseases are serious public health issues. Convergent Cross Mapping (CCM), the most popular method for causal detection in chaotic systems is known to result in high false detection rates in epidemiological settings, and a remedy to this problem is urgently sought. We introduce a method for refuting false causal relations based on a simple criterion of dimensionality, tested via the bootstrap. Our method gives excellent results on a two strain model, as well as on pneumonia and influenza data from the US. For both data sets CCM indeed generates high false detection rates. As a part of our investigation, we highlight the important role observation noise plays in the problem of causal detection. We then suggest a parameter free method to overcome observation noise in chaotic systems.
9_refuting_causal_relations_in_e.pdfPoster
Authors: Jie Kate Hu, Eric Tchetgen Tchetgen
Title: Causal Inference with Time Series Data and Unmeasured Confounding
Abstract: Unmeasured confounding threatens the validity of causal inference with observational studies. Recent proximal causal inference techniques show a new path tackling the problem by leveraging a pair of proxy variables of the unmeasured confounders U to estimate the causal effect. In this paper, we extend this approach to time series data. We first develop a method to estimate the causal effect of X on Y using proxies of U from a time series. We then develop a method to estimate the short-term causal effect in multi-site time series studies with a time-varying unmeasured confounder. Our approach allows the number of proxies to be many and weakly correlated with the unmeasured confounders, two issues often associated with time series data. We further adapted continuous updating generalized method of moment techniques to our inference procedure. We implemented an improved uncertainty metric for our causal effect estimates compared to off-the-shelf metrics from existing software. Finally, we validated our methods through simulation studies. These studies provide new insight into applying our method in practice.
10_causal_inference_with_time_ser.pdfPoster
Authors: Abigail Langbridge, Fearghal O'Donncha, Amadou Ba, Fabio Lorenzi, Christopher M. Lohse, Joern Ploennigs
Title: Can We Evaluate Causal Structures Using Graph Neural Networks?
Abstract: Causal models have significant potential to augment prediction models through improved interpretability and regularisation. However, their applicability is limited as they are computationally expensive, complex, and hard to verify particularly if dealing with large-scale, heterogeneous, non-stationary datasets. In this paper we present a framework for improving the accessibility of causal tools for large scale datasets through an analysis of decomposition and subsampling methods that we evaluate on the popular causal discovery method PCMCI+. Further, we propose a novel method for causal structure evaluation that utilises the regularisation effect of causal modelling to evaluate candidate causal structures on data without the need of ground-truth.
Scalable_Causal_Discovery_UAI_CamReady.pdf