Research Overview

My research focuses on the intersection between the formal study of causation and the fields of artificial intelligence (AI) and statistics in general, and machine learning in particular. This research area can broadly be structured into two intertwined threads: (I) causal inference for machine learning and (II) machine learning for causal inference.

Below, I outline the motivation and implications of each thread and contextualize my prior work within this framework.

Figure 1: Causal Inference Tasks Involved in My Research. Shaded/white nodes and ? denote observed/latent variables and quantities of interest, respectively. (a) Causal reasoning leverages known causal structure to predict effects of unseen interventions or distribution shifts, typically from purely observational data [13, 47]. (b) Causal discovery aims to infer a causal graph over observed variables [21, 48]. (c) In causal representation learning, the causal variables of interest are not directly observed and only accessible via low-level measurements such as pixels [49]. My work uses (a) to improve the robustness and explainability of AI and leverages machine learning methods for (b) and (c) from heterogeneous and high-dimensional data. Figure adapted from [50, 51].

(I) Causal Inference for Machine Learning

Despite remarkable progress, machine learning-based AI systems still face challenges that limit their use in high-stakes or safety-critical settings. On the one hand, they lack transparency and often reproduce or amplify historical biases in the data. On the other hand, they lack robustness and tend to generalize poorly to new domains. Our work on (I) seeks to address these challenges through causal reasoning (see Fig. 1a), which provides a powerful language for formalizing counterfactual notions such as fairness and explanations or analysing the effect of distribution shifts. Our work also highlights how viewing the underlying data-generating process from a causal perspective can help inform appropriate algorithm selection and inspire new assumptions and methods.

(I.A) Causal Reasoning for Algorithmic Recourse, Fairness, and Explainability

In AI-assisted consequential decision-making, algorithmic recourse [1] provides individuals who were unfavourably classified (e.g., not approved for a loan) with actionable recommendations on how to improve their situation. As this involves reasoning about the effect of interventions (e.g., lifestyle changes), it is fundamentally a causal problem [2]. We developed methods to identify low-cost interventions that achieve recourse with high probability [3–5] and proposed an individualized fairness criterion [6], which extends non-causal group-based notions of fair recourse [7] and is inspired by counterfactual fairness [8]. Counterfactual reasoning is also at the heart of providing explanations for why something occurred [9, 10]. Traditionally, such counterfactuals have been interpreted in interventional terms [11–13]. We formalized an alternative backtracking semantics [14] (Best Paper Award at CLeaR 2023), which is better suited for diagnostic reasoning tasks such as root cause analysis [15], and developed a practical implementation for causal systems consisting of deep generative models [16]. In a more applied project [17], we took a critical look at the inner workings of trajectory prediction methods by leveraging causal feature attribution methods [18, 19] and links to Granger causality [20].

(I.B) Implications of Causal Structure for Semi-Supervised and Transfer Learning

Causal mechanisms do not inform each other and tend to remain invariant when other parts of the system change [21]. The causal structure underlying a problem thus profoundly influences the effectiveness of machine learning approaches [22]. We studied such implications for covariate shift adaptation [23], semi-supervised learning [24], and natural language processing tasks [25], highlighting that cause and effect features should be treated differently, especially when learning from unlabelled data. When multiple source domains are available and the goal is to learn a predictor that perform well on new test domains, existing methods typically either optimise for average performance at the cost of reduced robustness [26, 27] or for worst-case performance by seeking invariant predictors that rely only on causal features [28, 29]. We have developed methods to interpolate between these objectives by learning predictors that generalize with high probability [30] and

to use invariant predictions as pseudo-labels to safely harness certain spurious (non-causal) features [31].

(I.C) Causality-Inspired Constraints for Representation Learning

In unsupervised representation learning, identifiability (i.e., provably inverting the data-generating process) requires restricting the model class [32, 33]. This often involves constraining the mixing function that maps latent variables to observations [34–36]. Inspired by the principle of independent causal mechanisms [21], we proposed an orthogonality condition on the columns of the Jacobian of the mixing function [37, 38], which facilitates blind source separation [39] and dimension reduction [40] while also explaining the effectiveness of VAEs [41] in representation learning [42]. For object-centric representation learning, we have drawn on causal generative scene models to develop methods [43, 44], as well as new assumptions that also impose specific structures on the Jacobian [45] or higher-order derivatives of the mixing function [46].

(II) Machine Learning for Causal Inference

Inferring causal structure or estimating causal effects from data often involves tasks such as regression, density estimation, conditional independence testing, or uncertainty quantification, which can be challenging for high-dimensional, nonlinear settings. Our work on (II) leverages machine learning to extend, improve, and develop new causal inference methods for such scenarios. Beyond classical causal discovery (see Fig. 1b), a particular focus has been on the theoretical foundations of causal representation learning (see Fig. 1c).

(II.A) Merging Causal Insights from Different Sources

In the age of big data, effectively integrating knowledge from different sources is promising but non-trivial [52]. To this end, we developed an optimisation-based approach for merging causal insights from datasets sharing a subset of variables [53] and a biased but lower-variance causal effect estimator that combines observational and interventional data [54], inspired by ideas in shrinkage estimation [55, 56].

(II.B) Causal Discovery from Heterogeneous or Experimental Data

Since correlation does not imply causation, discerning causal structure is notoriously difficult from observational data and often requires additional assumptions or heterogeneous (non-i.i.d.) observations. Given non-identically distributed datasets, we proposed seeking causal graphs that provide parsimonious explanations by minimizing the number of mechanism shifts across domains [57]. This yields a principled method with theoretical guarantees [58] and flexible estimators, e.g., via kernel-based conditional independence tests [59], which we also investigated for non-stationary or functional data [60, 61]. Beyond mere heterogeneity, data resulting from direct experimentation is even more valuable for causal discovery but typically also costly to collect, motivating careful design strategies. To this end, we developed a Bayesian active learning framework, in which the current beliefs about the underlying causal model are used to inform which subsequent intervention would provide the most information about the causal query of interest (e.g., a particular treatment effect or edge in the graph) [62, 63]. To facilitate tractable approximate inference, this framework leverages probabilistic machine learning methods such as Gaussian process regression [64], Bayesian optimization [65], and variational inference [66].

(II.C) Identifiable Causal Representation Learning

To overcome the obstacle that high-level variables of interest (e.g., properties of objects in a scene) are often not directly accessible, causal representation learning aims to integrate causal discovery with unsupervised learning [67, 68] to infer latent variable models endowed with causal structure. To understand under which assumptions and up what ambiguities learning such latent causal models is possible, identifiability studies are crucial. In our work, we have characterized sufficient conditions for identifiability across several settings [69], including when learning from multiple non-independent views sharing some of the same latent variables, arising, e.g., from data augmentation [70, 71] or different modalities [72]; multiple non-identically distributed datasets, related via interventions in a shared causal model [51, 73]; or i.i.d. data, subject to linearity and sparsity constraints [74]. These theoretical results often also suggest suitable estimation procedures, e.g., based on self-supervised learning [75] or deep generative models such as normalizing flows [76], endowed with latent causal structure.

Applications

During the Covid-19 pandemic, we carried out a causal mediation analysis of case fatality rates across different countries by leveraging demographic data to separate age-specific and other effects [77] and compared counterfactual vaccination strategies in Israel by combining causal models with epidemiological simulations and literature estimates of key parameters [78]. Further, in collaboration with researchers from Imperial College London, we sought to better understand the interlinkages between the UN sustainable development goals and climate change [79, 80].

CURRENT AND FUTURE RESEARCH

My previous work demonstrates the potential synergies of integrating causality with machine learning, offering new insights and tools for both fields. In my current and future research, I intend to continue studying theoretical and methodological questions at this intersection to further enhance the transparency and domain-shift robustness of AI systems and extend the applicability of causal inference methods to more complex real-world data. Additionally, I am eager to apply these methods to scientific challenges in other disciplines, with a particular focus on single-cell biology—a field that I believe presents unique and exciting opportunities for causal machine learning.

References

[1] B. Ustun, A. Spangher, and Y. Liu, “Actionable recourse in linear classification,” in Conference on Fairness, Accountability, and Transparency, pp. 10–19, 2019.[2] A.-H. Karimi, B. Schölkopf, and I. Valera, “Algorithmic recourse: from counterfactual explanations to interventions,” in Conference on Fairness, Accountability, and Transparency, pp. 353–362, 2021.[3] A.-H. Karimi, J. von Kügelgen, B. Schölkopf, and I. Valera, “Algorithmic recourse under imperfect causal knowledge: a probabilistic approach,” in Advances in Neural Information Processing Systems, vol. 33, pp. 265–277, 2020.[4] J. von Kügelgen, N. Agarwal, J. Zeitler, A. Mastouri, and B. Schölkopf, “Algorithmic recourse in partially and fully confounded settings through bounding counterfactual effects,” in ICML Workshop on Algorithmic Recourse, 2021.[5] A.-H. Karimi, J. von Kügelgen, B. Schölkopf, and I. Valera, “Towards causal algorithmic recourse,” in xxAI - Beyond Explainable AI. Lecture Notes in Computer Science, vol. 13200, pp. 139–166, Springer, 2022.[6] J. von Kügelgen, A.-H. Karimi, U. Bhatt, I. Valera, A. Weller, and B. Schölkopf, “On the fairness of causal algorithmic recourse,” in AAAI Conference on Artificial Intelligence, vol. 36, pp. 9584–9594, 2022.[7] V. Gupta, P. Nokhiz, C. D. Roy, and S. Venkatasubramanian, “Equalizing recourse across groups,” arXiv:1909.03166, 2019.[8] M. J. Kusner, J. Loftus, C. Russell, and R. Silva, “Counterfactual fairness,” in Advances in Neural Information Processing Systems, vol. 30, 2017.[9] S. Beckers, “Causal explanations and XAI,” in Conference on Causal Learning and Reasoning, pp. 90–109, PMLR, 2022.[10] S. Wachter, B. Mittelstadt, and C. Russell, “Counterfactual explanations without opening the black box: automated decisions and the GDPR,” Harvard Journal of Law and Technology, vol. 31, no. 2, 2018.[11] D. Lewis, Counterfactuals. Blackwell Publishers, 1973.[12] D. Lewis, “Counterfactual dependence and time’s arrow,” Nous, pp. 455–476, 1979.[13] J. Pearl, Causality: Models, Reasoning, and Inference. Cambridge University Press, 2009.[14] J. von Kügelgen, A. Mohamed, and S. Beckers, “Backtracking counterfactuals,” in Conference on Causal Learning and Reasoning, pp. 177–196, PMLR, 2023.[15] K. Budhathoki, L. Minorics, P. Blöbaum, and D. Janzing, “Causal structure-based root cause analysis of outliers,” in International Conference on Machine Learning, pp. 2357–2369, PMLR, 2022.[16] K.-R. Kladny, J. von Kügelgen, B. Schölkopf, and M. Muehlebach, “Deep backtracking counterfactuals for causally compliant explanations,” Transactions on Machine Learning Research, 2024.[17] O. Makansi, J. von Kügelgen, F. Locatello, P. V. Gehler, D. Janzing, T. Brox, and B. Schölkopf, “You mostly walk alone: Analyzing feature attribution in trajectory prediction,” in International Conference on Learning Representations, 2022.[18] L. S. Shapley, “A value for n-person games,” Contribution to the Theory of Games, vol. 2, 1953.[19] D. Janzing, L. Minorics, and P. Bloebaum, “Feature relevance quantification in explainable AI: A causal problem,” in International Conference on Artificial Intelligence and Statistics, pp. 2907–2916, PMLR, 2020.[20] C. W. Granger, “Investigating causal relations by econometric models and cross-spectral methods,” Econometrica: Journal of the Econometric Society, pp. 424–438, 1969.[21] J. Peters, D. Janzing, and B. Schölkopf, Elements of Causal Inference: Foundations and Learning Algorithms. MIT Press, 2017.[22] B. Schölkopf, D. Janzing, J. Peters, E. Sgouritsa, K. Zhang, and J. Mooij, “On causal and anticausal learning,” in International Conference on Machine Learning, pp. 459–466, 2012.[23] J. von Kügelgen, A. Mey, and M. Loog, “Semi-generative modelling: Covariate-shift adaptation with cause and effect features,” in International Conference on Artificial Intelligence and Statistics, pp. 1361–1369, PMLR, 2019.[24] J. von Kügelgen, A. Mey, M. Loog, and B. Schölkopf, “Semi-supervised learning, causality, and the conditional cluster assumption,” in Conference on Uncertainty in Artificial Intelligence, pp. 1–10, PMLR, 2020.[25] Z. Jin, J. von Kügelgen, J. Ni, T. Vaidhya, A. Kaushal, M. Sachan, and B. Schölkopf, “Causal direction of data collection matters: Implications of causal and anticausal learning for NLP,” in Conference on Empirical Methods in Natural Language Processing, pp. 9499–9513, 2021.[26] K. Muandet, D. Balduzzi, and B. Schölkopf, “Domain generalization via invariant feature representation,” in International Conference on Machine Learning, pp. 10–18, PMLR, 2013.[27] G. Blanchard, G. Lee, and C. Scott, “Generalizing from several related classification tasks to a new unlabeled sample,” in Advances in Neural Information Processing Systems, vol. 24, pp. 2178–2186, 2011.[28] M. Rojas-Carulla, B. Schölkopf, R. Turner, and J. Peters, “Invariant models for causal transfer learning,” Journal of Machine Learning Research, vol. 19, no. 36, pp. 1–34, 2018.[29] M. Arjovsky, L. Bottou, I. Gulrajani, and D. Lopez-Paz, “Invariant risk minimization,” arXiv preprint arXiv:1907.02893, 2019.[30] C. Eastwood, A. Robey, S. Singh, J. von Kügelgen, H. Hassani, G. J. Pappas, and B. Schölkopf, “Probable domain generalization via quantile risk minimization,” in Advances in Neural Information Processing Systems, vol. 35, pp. 17340–17358, 2022.[31] C. Eastwood, S. Singh, A. L. Nicolicioiu, M. Vlastelica, J. von Kügelgen, and B. Schölkopf, “Spuriosity didn’t kill the classifier: using invariant predictions to harness spurious features,” in Advances in Neural Information Processing Systems, pp. 18291–18324, 2023.[32] A. Hyvärinen and P. Pajunen, “Nonlinear independent component analysis: Existence and uniqueness results,” Neural networks, 1999.[33] F. Locatello, S. Bauer, M. Lucic, G. Raetsch, S. Gelly, B. Schölkopf, and O. Bachem, “Challenging common assumptions in the unsupervised learning of disentangled representations,” in International Conference on Machine Learning, 2019.[34] P. Comon, “Independent component analysis, a new concept?,” Signal processing, vol. 36, no. 3, pp. 287–314, 1994.[35] K. Zhang and A. Hyvärinen, “On the identifiability of the post-nonlinear causal model,” in Conference on Uncertainty in Artificial Intelligence, pp. 647–655, 2009.[36] A. Immer, C. Schultheiss, J. E. Vogt, B. Schölkopf, P. Bühlmann, and A. Marx, “On the identifiability and estimation of causal location-scale noise models,” in International Conference on Machine Learning, pp. 14316–14332, PMLR, 2023.[37] L. Gresele, J. von Kügelgen, V. Stimper, B. Schölkopf, and M. Besserve, “Independent mechanism analysis, a new concept?,” in Advances in Neural Information Processing Systems, vol. 34, pp. 28233–28248, 2021.[38] S. Ghosh, L. Gresele, J. von Kügelgen, M. Besserve, and B. Schölkopf, “Independent mechanism analysis and the manifold hypothesis,” in NeurIPS Causal Representation Learning Workshop, 2023.[39] S. Buchholz, M. Besserve, and B. Schölkopf, “Function classes for identifiable nonlinear independent component analysis,” Advances in Neural Information Processing Systems, vol. 35, pp. 16946–16961, 2022.[40] E. Cunningham, A. D. Cobb, and S. Jha, “Principal component flows,” in International Conference on Machine Learning, pp. 4492–4519, PMLR, 2022.[41] D. P. Kingma and M. Welling, “Auto-encoding variational bayes,” in International Conference on Learning Representations, 2014.[42] P. Reizinger, L. Gresele, J. Brady, J. von Kügelgen, D. Zietlow, B. Schölkopf, G. Martius, W. Brendel, and M. Besserve, “Embrace the gap: VAEs perform independent mechanism analysis,” in Advances in Neural Information Processing Systems, vol. 35, pp. 12040–12057, 2022.[43] J. von Kügelgen, I. Ustyuzhaninov, P. Gehler, M. Bethge, and B. Schölkopf, “Towards causal generative scene models via competition of experts,” in ICLR Workshop on Causal Learning for Decision Making, 2020.[44] M. Tangemann, S. Schneider, J. von Kügelgen, F. Locatello, P. Gehler, T. Brox, M. Kümmerer, M. Bethge, and B. Schölkopf, “Unsupervised object learning via common fate,” in Conference on Causal Learning and Reasoning, pp. 281–327, PMLR, 2023.[45] J. Brady, R. S. Zimmermann, Y. Sharma, B. Schölkopf, J. von Kügelgen, and W. Brendel, “Provably learning object-centric representations,” in International Conference on Machine Learning, pp. 3038–3062, PMLR, 2023. [46] J. Brady, J. von Kügelgen, S. Lachapelle, S. Buchholz, and W. Brendel, “Interaction asymmetry: A general principle for learning composable abstractions,” in NeurIPS Causal Representation Learning Workshop, 2024.[47] B. Schölkopf, “Causality for machine learning,” in Probabilistic and Causal Inference: The Works of Judea Pearl, pp. 765–804, ACM, 2022.[48] P. Spirtes, C. Glymour, and R. Scheines, Causation, Prediction, and Search. MIT press, 2001.[49] B. Schölkopf, F. Locatello, S. Bauer, N. R. Ke, N. Kalchbrenner, A. Goyal, and Y. Bengio, “Toward causal representation learning,” Proceedings of the IEEE, vol. 109, no. 5, pp. 612–634, 2021.[50] B. Schölkopf and J. von Kügelgen, “From statistical to causal learning,” in Proceedings of the International Congress of Mathematicians, vol. 7, pp. 5540–5593, 2022. [51] J. von Kügelgen, M. Besserve, L. Wendong, L. Gresele, A. Kekic, E. Bareinboim, D. M. Blei, and B. Schölkopf, “Nonparametric identifiability of causal representations from unknown interventions,” in Advances in Neural Information Processing Systems, pp. 48603–48638, 2023.[52] E. Bareinboim and J. Pearl, “Causal inference and the data-fusion problem,” Proceedings of the National Academy of Sciences, vol. 113, no. 27, pp. 7345–7352, 2016.[53] L. Gresele, J. von Kügelgen, J. Kübler, E. Kirschbaum, B. Schölkopf, and D. Janzing, “Causal inference through the structural causal marginal problem,” in International Conference on Machine Learning, pp. 7793–7824, PMLR, 2022. [54] K.-R. Kladny, J. von Kügelgen, B. Schölkopf, and M. Muehlebach, “Causal effect estimation from observational and interventional data through matrix weighted linear estimators,” in Conference on Uncertainty in Artificial Intelligence, pp. 1087–1097, PMLR, 2023.[55] W. James and C. Stein, “Estimation with quadratic loss,” in Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Contributions to the Theory of Statistics, vol. 4, pp. 361–380, University of California Press, 1961.[56] R. Tibshirani, “Regression shrinkage and selection via the lasso,” Journal of the Royal Statistical Society Series B: Statistical Methodology, vol. 58, no. 1, pp. 267–288, 1996.[57] R. Perry, J. von Kügelgen, and B. Schölkopf, “Causal discovery in heterogeneous environments under the sparse mechanism shift hypothesis,” in Advances in Neural Information Processing Systems, vol. 35, pp. 10904–10917, 2022.[58] A. Jaber, M. Kocaoglu, K. Shanmugam, and E. Bareinboim, “Causal discovery from soft interventions with unknown targets: Characterization and learning,” Advances in Neural Information Processing Systems, vol. 33, pp. 9551–9561, 2020.[59] K. Zhang, J. Peters, D. Janzing, and B. Schölkopf, “Kernel-based conditional independence test and application in causal discovery,” in Conference on Uncertainty in Artificial Intelligence, pp. 804–813, 2011.[60] F. Laumann, J. von Kügelgen, and M. Barahona, “Kernel two-sample and independence tests for nonstationary random processes,” in Engineering Proceedings, vol. 5, p. 31, MDPI, 2021.[61] F. Laumann, J. von Kügelgen, J. Park, B. Schölkopf, and M. Barahona, “Kernel-based independence tests for causal structure learning on functional data,” Entropy, vol. 25, no. 12, p. 1597, 2023.[62] J. von Kügelgen, P. K. Rubenstein, B. Schölkopf, and A. Weller, “Optimal experimental design via Bayesian optimization: active causal structure learning for Gaussian process networks,” in NeurIPS Workshop “Do the right thing”: machine learning and causal inference for improved decision making, 2019.[63] C. Toth, L. Lorch, C. Knoll, A. Krause, F. Pernkopf, R. Peharz, and J. von Kügelgen, “Active Bayesian causal inference,” in Advances in Neural Information Processing Systems, vol. 35, pp. 16261–16275, 2022.[64] C. K. Williams and C. E. Rasmussen, Gaussian Processes for Machine Learning. MIT Press, 2006.[65] J. Snoek, H. Larochelle, and R. P. Adams, “Practical Bayesian optimization of machine learning algorithms,” in Advances in Neural Information Processing Systems, vol. 25, 2012.[66] D. M. Blei, A. Kucukelbir, and J. D. McAuliffe, “Variational inference: A review for statisticians,” Journal of the American Statistical Association, vol. 112, no. 518, pp. 859–877, 2017.[67] Y. Bengio, A. Courville, and P. Vincent, “Representation learning: A review and new perspectives,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 35, no. 8, pp. 1798–1828, 2013.[68] A. Hyvärinen, I. Khemakhem, and H. Morioka, “Nonlinear independent component analysis for principled disentanglement in unsupervised deep learning,” Patterns, vol. 4, no. 10, 2023.[69] J. von Kügelgen, Identifiable Causal Representation Learning: Unsupervised, Multi-View, and Multi-Environment. PhD thesis, University of Cambridge, 2024.[70] J. von Kügelgen, Y. Sharma, L. Gresele, W. Brendel, B. Schölkopf, M. Besserve, and F. Locatello, “Self-supervised learning with data augmentations provably isolates content from style,” in Advances in Neural Information Processing Systems, vol. 34, pp. 16451–16467, 2021.[71] C. Eastwood, J. von Kügelgen, L. Ericsson, D. Bouchacourt, P. Vincent, M. Ibrahim, and B. Schölkopf, “Self-supervised disentanglement by leveraging structure in data augmentations,” in NeurIPS Causal Representation Learning Workshop, 2023.[72] D. Yao, D. Xu, S. Lachapelle, S. Magliacane, P. Taslakian, G. Martius, J. von Kügelgen, and F. Locatello, “Multi-view causal representation learning with partial observability,” in International Conference on Learning Representations, 2024.[73] L. Wendong, A. Kekic, J. von Kügelgen, S. Buchholz, M. Besserve, L. Gresele, and B. Schölkopf, “Causal component analysis,” in Advances in Neural Information Processing Systems, pp. 32481–32520, 2023.[74] D. Xu, D. Yao, S. Lachapelle, P. Taslakian, J. von Kügelgen, F. Locatello, and S. Magliacane, “A sparsity principle for partially observable causal representation learning,” in International Conference on Machine Learning, 2024.[75] T. Chen, S. Kornblith, M. Norouzi, and G. Hinton, “A simple framework for contrastive learning of visual representations,” in International Conference on Machine Learning, pp. 1597–1607, 2020.[76] G. Papamakarios, E. Nalisnick, D. J. Rezende, S. Mohamed, and B. Lakshminarayanan, “Normalizing flows for probabilistic modeling and inference,” Journal of Machine Learning Research, vol. 22, no. 57, pp. 1–64, 2021.[77] J. von Kügelgen, L. Gresele, and B. Schölkopf, “Simpson’s paradox in COVID-19 case fatality rates: a mediation analysis of age-related causal effects,” IEEE Transactions on Artificial Intelligence, vol. 2, no. 1, pp. 18–27, 2021.[78] A. Kekic, J. Dehning, L. Gresele, J. von Kügelgen, V. Priesemann, and B. Schölkopf, “Evaluating vaccine allocation strategies using simulation-assisted causal modelling,” Patterns, vol. 4, no. 6, 2023.[79] F. Laumann, J. von Kügelgen, and M. Barahona, “Non-linear interlinkages and key objectives amongst the Paris agreement and the sustainable development goals,” in ICLR Workshop: Tackling Climate Change with Machine Learning, 2020.[80] F. Laumann, J. von Kügelgen, T. H. K. Uehara, and M. Barahona, “Complex interlinkages, key objectives, and nexuses among the sustainable development goals and climate change: a network analysis,” The Lancet Planetary Health, vol. 6, no. 5, pp. e422–e430, 2022.

Page updated

Google Sites

Report abuse