Skip to main content

Resultados relacionados

Resultados totales (Incluyendo duplicados): 5
Encontrada(s) 1 página(s)

e-Archivo. Repositorio Institucional de la Universidad Carlos III de Madrid
Publication Artículo científico (JournalArticle)

Explainable Earth Surface Forecasting Under Extreme Events

  • Pellicer Valero, Oscar José
  • Fernández Torres, Miguel Ángel
  • Ji, Chaonan
  • Mahecha, Miguel D.
  • Camps Valls, Gustau
With climate change-related extreme events on the rise, high-dimensional Earth observation data present a unique opportunity for forecasting and understanding impacts on ecosystems. This is, however, impeded by the complexity of processing, visualizing, modeling, and explaining this data. We train a convolutional long short-term memory-based architecture on the novel DeepExtremeCubes data set to showcase how this challenge can be met. DeepExtremeCubes includes around 40,000 long-term Sentinel-2 minicubes (January 2016–October 2022) worldwide, along with labeled extreme events, meteorological data, vegetation land cover, and a topography map, sampled from locations affected by extreme climate events and surrounding areas. When predicting future reflectances and vegetation impacts through the kernel normalized difference vegetation index, the model achieved an R2 score of 0.9055 in the test set. Explainable artificial intelligence was used to analyze the model's predictions during the October 2020 Central South America compound heatwave and drought event. We chose the same area exactly 1 year before the event as a counterfactual, finding that the average temperature and surface pressure are generally the most important predictors. In contrast, minimum evaporation anomalies play a leading role during the event. We also found the anomalies of the reflectances in the timestep before the extreme event to be critical predictors of its impact on vegetation. The code to replicate all experiments and figures in this paper is publicly available at https://github.com/DeepExtremes/txyXAI., This work was supported by the ESA AI4Science project Multi-Hazards, Compounds and Cascade events: DeepExtremes, 2022-2024; the European Union's Horizon 2020 research and innovation project XAIDA: Extreme Events Artificial Intelligence for Detection and Attribution, 2021-2024 (Grant agreement No 101003469); and the European Union's Horizon 2022 project ELIAS: European Lighthouse of AI for Sustainability, 2023-2027 (Grant agreement No. 101120237). The authors acknowledge the support from the computer resources provided by Artemisa (funded by the European Union ERDF and Comunitat Valenciana), as well as the technical support provided by the Instituto de Física Corpuscular, IFIC (CSIC-UV).
RUA. Repositorio Institucional de la Universidad de Alicante
Publication Otros (Other)

Privacy and Accuracy Implications of Model Complexity and Integration in Heterogeneous Federated Learning

  • Németh, Gergely Dániel
  • Lozano, Miguel Angel
  • Quadrianto, Novi
  • Oliver, Nuria
  • Lozano, Miguel Angel|||/items/413f111d-7a63-43f2-bb70-6421affe1647
Federated Learning (FL) has been proposed as a privacy-preserving solution for distributed machine learning, particularly in heterogeneous FL settings where clients have varying computational capabilities and thus train models with different complexities compared to the server’s model. However, FL is not without vulnerabilities: recent studies have shown that it is susceptible to membership inference attacks (MIA), which can compromise the privacy of client data. In this paper, we examine the intersection of these two aspects, heterogeneous FL and its privacy vulnerabilities, by focusing on the role of client model integration, the process through which the server integrates parameters from clients’ smaller models into its larger model. To better understand this process, we first propose a taxonomy that categorizes existing heterogeneous FL methods and enables the design of seven novel heterogeneous FL model integration strategies. Using CIFAR-10, CIFAR-100, and FEMNIST vision datasets, we evaluate the privacy and accuracy trade-offs of these approaches under three types of MIAs. Our findings reveal significant differences in privacy leakage and performance depending on the integration method. Notably, introducing randomness in the model integration process enhances client privacy while maintaining competitive accuracy for both the clients and the server. This work provides quantitative light on the privacy-accuracy implications client model integration in heterogeneous FL settings, paving the way towards more secure and efficient FL systems.
RUA. Repositorio Institucional de la Universidad de Alicante
Publication Otros (Other)

What is beautiful is still good: the attractiveness halo effect in the era of beauty filters

  • Gulati, Aditya
  • Martínez-Garcia, Marina
  • Fernández, Daniel
  • Lozano, Miguel Angel
  • Lepri, Bruno
  • Oliver, Nuria
  • Lozano, Miguel Angel|||/items/413f111d-7a63-43f2-bb70-6421affe1647
The impact of cognitive biases on decision-making in the digital world remains under-explored despite its well-documented effects in physical contexts. This paper addresses this gap by investigating the attractiveness halo effect using AI-based beauty filters. We conduct a large-scale online user study involving 2748 participants who rated facial images from a diverse set of 462 distinct individuals in two conditions: original and attractive after applying a beauty filter. Our study reveals that the same individuals receive statistically significantly higher ratings of attractiveness and other traits, such as intelligence and trustworthiness, in the attractive condition. We also study the impact of age, gender and ethnicity and identify a weakening of the halo effect in the beautified condition, resolving conflicting findings from the literature and suggesting that filters could mitigate this cognitive bias. Finally, our findings raise ethical concerns regarding the use of beauty filters.
RUA. Repositorio Institucional de la Universidad de Alicante
Publication Otros (Other)

Relaxing Core Assumptions: the Impact of Data, Model and Participation Heterogeneity on Performance, Privacy and Fairness in Federated Learning

  • Németh, Gergely Dániel
Federated Learning (FL) enables decentralized training of machine learning models on distributed data while preserving privacy by design. An FL design consists of clients training models on private data and a central server aggregating a global model based on the consensus among clients. In an ideal scenario, the training data and computing resources are identically and independently distributed (i.i.d.) among clients, therefore, clients can work together in agreement to reach a global optima. However, in realistic FL settings, heterogeneity arises between clients in terms of both data and resource availability. This research focuses on such scenarios, with a special interest on how the server can adapt the aggregation method from a simple averaging to address the clients’ diversity. The first research direction discusses existing client selection methods and proposes a novel taxonomy of FL methods where the participation of the clients is actively managed by the server to achieve a global objective with respect to the client heterogeneity. This research direction is presented in [NLQO22]. The next chapter focuses on model heterogeneity as an inclusion policy for low-resource clients. It investigates the implications of client resource constraints on privacy given a reduced model complexity in low-resource clients. This work has been presented in [NLQO25]. The final area provides a solution to the data heterogeneity problem with distribution aware client selection. Applying this solution can mitigate spurious correlations and improve algorithmic fairness in FL. This research line has been described in [NFN+25]. [NLQO22] Németh, G. D., Lozano, M. A., Quadrianto, N., and Oliver, N. (2022). A Snapshot of the Frontiers of Client Selection in Federated Learning. Transactions on Machine Learning Research. [NLQO25] Németh, G. D., Lozano, M. A., Quadrianto, N., and Oliver, N. (2025). Privacy and Accuracy Implications of Model Complexity and Integration in Heterogeneous Federated Learning. IEEE Access, 13, 40258-40274. [NFN+25] Németh, G. D., Fani, E., Ng, Y. J., Caputo, B., Lozano, M. A., Oliver, N., and Quadrianto, N.(2025). FedDiverse: Tackling Data Heterogeneity in Federated Learning with Diversity-Driven Client Selection.
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Linear quadratic control of nonlinear systems with Koopman operator learning and the Nyström method

  • Caldarelli, Edoardo
  • Chatalic, Antoine
  • Colomé, Adrià
  • Molinari, Cesare
  • Ocampo-Martínez, Carlos
  • Torras, Carme
  • Rosasco, Lorenzo
In this paper, we study how the Koopman operator framework can be combined with kernel methods to effectively control nonlinear dynamical systems. While kernel methods have typically large computational requirements, we show how random subspaces (Nystr\"om approximation) can be used to achieve huge computational savings while preserving accuracy. Our main technical contribution is deriving theoretical guarantees on the effect of the Nystr\"om approximation. More precisely, we study the linear quadratic regulator problem, showing that the approximated Riccati operator converges at the rate $m^{-1/2}$, and the regulator objective, for the associated solution of the optimal control problem, converges at the rate $m^{-1}$, where $m$ is the random subspace size. Theoretical findings are complemented by numerical experiments corroborating our results., E. Caldarelli, A. Colomé and C. Torras acknowledge support from the project CLOTHILDE (“CLOTH manIpulation Learning from DEmonstrations”), funded by the European Research Council(ERC) under the European Union’s Horizon 2020 research and innovation programme (Advanced Grant agreement No 741930). E. Caldarelli acknowledges travel support from ELISE (GA No 951847). C. Molinari is part of “GNAMPA” (INdAM) and has been supported by the projects MIUR Excellence Department awarded to DIMA UniGe CUP D33C23001110001, AFOSR FA8655-22-1-7034, MIUR-PRIN 202244A7YL and PON “Ricerca e Innovazione” 2014–2020. C. Ocampo-Martinez acknowledges the support from the project SEAMLESS (PID2023-148840OB-I00) funded by MCIN/AEI. L. Rosasco acknowledges the financial support of the European Research Council, (grant SLING 819789), the European Commission, (Horizon Europe grant ELIAS 101120237), the US Air Force Office of Scientific Research (FA8655-22-1-7034), the Ministry of Education, University and Research (FARE grant ML4IP R205T7J2KP; grant BAC FAIR PE00000013 funded by the EU - NGEU) and the Center for Brains, Minds and Machines (CBMM) ., Peer reviewed