Santiago Zazo

Accurate survival prediction is critical in oncology for prognosis and treatment planning. Traditional approaches often rely on a single data modality, limiting their ability to capture the complexity of tumor biology. To address this challenge, we introduce a multimodal deep learning framework for survival analysis capable of modeling both single and competing risks scenarios, evaluating the impact of integrating multiple medical data sources on survival predictions. We propose SAMVAE (Survival Analysis Multimodal Variational Autoencoder), a novel deep learning architecture designed for survival prediction that integrates six data modalities: clinical variables, four molecular profiles, and histopathological images. SAMVAE leverages modality-specific encoders to project inputs into a shared latent space, enabling robust survival prediction while preserving modality-specific information. We evaluate SAMVAE on two cancer cohorts—breast cancer and lower-grade glioma—applying tailored preprocessing, dimensionality reduction, and hyperparameter optimization. The results demonstrate the successful integration of multimodal data for both standard survival analysis and competing risks scenarios across different datasets. Our model achieves competitive predictive performance compared to state-of-the-art multimodal survival models. To the best of our knowledge, among publicly available implementations, this framework represents an early approach to integrating clinical, multi-omics, and imaging data for continuous-time competing risks. SAMVAE introduces a probabilistic framework that supports the generation of personalized survival curves from heterogeneous data sources. Its parametric formulation enables the derivation of clinically meaningful statistics from the output distributions, providing patient-specific insights through interactive multimedia. These properties position it as a methodological tool with the potential to support risk modeling and interpretable survival analysis in oncology. Importantly, its main contribution is a unified probabilistic multimodal fusion framework.

Accurate underwater localization is crucial for autonomous underwater vehicles, yet it remains challenging due to the complex and dynamic nature of marine environments. Traditional localization methods that rely on anchor nodes and precise synchronization often fail to adapt to real-world underwater acoustic channel conditions characterized by multipath effects, high attenuation, and significant variability. Recent advances in deep learning offer model-free approaches that mitigate some of these issues, but they still struggle with generalization in scenarios with limited data. In this study, we introduce the application of attentive neural processes (ANPs) to underwater localization, enhancing both accuracy and robustness through effective uncertainty modeling and dynamic adaptation based on observed context data. By leveraging the ANP framework, our approach paves the way for extensive offline pretraining on simulated data, followed by rapid few-shot adaptation under real conditions; in this work, we evaluate near-zero-shot generalization. This strategy not only improves predictive performance but also significantly reduces the need for large-scale real-world data, making it highly suitable for dynamic underwater environments. Extensive evaluations demonstrate that ANPs significantly outperform traditional multilayer perceptron approaches, achieving reductions in mean absolute error of up to 25% compared to standard baselines. These results highlight the potential of ANPs to deliver reliable, adaptable, and precise localization solutions in challenging underwater scenarios.

Deep neural networks achieve state-of-the-art performance in estimating heterogeneous treatment effects, but their opacity limits trust and adoption in sensitive domains such as medicine, economics and public policy. Building on well-established and high-performing causal neural architectures, we propose causalKANs, a framework that transforms neural estimators of conditional average treatment effects(CATEs) into Kolmogorov–Arnold Networks (KANs). By incorporating pruning and symbolic simplification, causalKANsyields interpretable closed-form formulas while preserving predictive accuracy. Experiments on benchmark datasets demonstrate that causalKANs perform on par with neural baselines in CATE error metrics, and that even simple KAN variants achieve competitive performance, offering a favorable accuracy–interpretability trade-off. By combining reliability with analytic accessibility, causalKANs provide auditable estimators supported by closed-form expressions and interpretable plots, enabling trustworthy individualized decision-making in high-stakes settings. We release the code for reproducibility.

Why should a clinician trust an Artificial Intelligence (AI) prediction? Despite the increasing accuracy of machine learning methods in medicine, the lack of transparency continues to hinder their adoption in clinical practice. In this work, we explore Kolmogorov-Arnold Networks (KANs) for clinical classification tasks on tabular data. In contrast to traditional neural networks, KANs are function-based architectures that offer intrinsic interpretability through transparent, symbolic representations. We introduce \emph{Logistic-KAN}, a flexible generalization of logistic regression, and \emph{Kolmogorov-Arnold Additive Model (KAAM)}, a simplified additive variant that delivers transparent, symbolic formulas. Unlike ``black-box'' models that require post-hoc explainability tools, our models support built-in patient-level insights, intuitive visualizations, and nearest-patient retrieval. Across multiple health datasets, our models match or outperform standard baselines, while remaining fully interpretable. These results position KANs as a promising step toward trustworthy AI that clinicians can understand, audit, and act upon. We release the code for reproducibility in our repository.

Underwater vessel localization using passive magnetic anomaly sensing is a challenging problem due to the variability in vessel magnetic signatures and operational conditions. Data-based approaches may fail to generalize even to slightly different conditions. Thus, we propose an Attentive Neural Process (ANP) approach, in order to take advantage of its few-shot capabilities to generalize, for robust localization of underwater vessels based on magnetic anomaly measurements. Our ANP models the mapping from multi-sensor magnetic readings to position as a stochastic function: it cross-attends to a variable-size set of context points and fuses these with a global latent code that captures trajectory-level factors. The decoder outputs a Gaussian over coordinates, providing both point estimates and well-calibrated predictive variance. We validate our approach using a comprehensive dataset of magnetic disturbance fields, covering 64 distinct vessel configurations (combinations of varying hull sizes, submersion depths (water-column height over a seabed array), and total numbers of available sensors). Six magnetometer sensors in a fixed circular arrangement record the magnetic field perturbations as a vessel traverses sinusoidal trajectories. We compare the ANP against baseline multilayer perceptron (MLP) models: (1) base MLPs trained separately on each vessel configuration, and (2) a domain-randomized search (DRS) MLP trained on the aggregate of all configurations to evaluate generalization across domains. The results demonstrate that the ANP achieves superior generalization to new vessel conditions, matching the accuracy of configuration-specific MLPs while providing well-calibrated uncertainty quantification. This uncertainty-aware prediction capability is crucial for real-world deployments, as it can inform adaptive sensing and decision-making. Across various in-distribution scenarios, the ANP halves the mean absolute error versus a domain-randomized MLP (0.43 m vs. 0.84 m). The model is even able to generalize to out-of-distribution data, which means that our approach has the potential to facilitate transferability from offline training to real-world conditions.

Synthetic Data Generation (SDG) is a promising solution for healthcare, offering the potential to generate synthetic patient data closely resembling real-world data while preserving privacy. However, data scarcity and heterogeneity, particularly in under-resourced regions, challenge the effective implementation of SDG. This paper addresses these challenges using Federated Learning (FL) for SDG, focusing on sharing synthetic patients across nodes. By leveraging collective knowledge and diverse data distributions, we hypothesize that sharing synthetic data can significantly enhance the quality and representativeness of generated data, particularly for institutions with limited or biased datasets. This approach aligns with meta-learning concepts, like Domain Randomized Search. We compare two FL techniques, FedAvg and Synthetic Data Sharing (SDS), the latter being our proposed contribution. Both approaches are evaluated using variational autoencoders with Bayesian Gaussian mixture models across diverse medical datasets. Our results demonstrate that while both methods improve SDG, SDS consistently outperforms FedAvg, producing higher-quality, more representative synthetic data. Non-IID scenarios reveal that while FedAvg achieves improvements of 13–27% in reducing divergence compared to isolated training, SDS achieves reductions exceeding 50% in the worst-performing nodes. These findings underscore synthetic data sharing potential to reduce disparities between data-rich and data-poor institutions, fostering more equitable healthcare research and innovation.

In recent years, passive vehicle safety has become one of the major concerns for the automotive industry due to the considerable increase in the use of cars as a means of daily transport. Since real crash testing has a high financial cost, finite element simulations are generally used, which entail high computational cost and long simulation times. In this paper, we make use of the recent advances in the deep learning field to propose an affordable method to provide reliable approximations of the finite element simulator model that significantly reduce the computational load and time required. We compare the prediction performance in crash tests of different models, namely feed-forward neural networks and bayesian neural networks, as well as two multi-output regression methods. Our results show promising results, as deep learning models are able to drastically reduce the engineering costs while providing a feasible first approximation to the passenger’s injuries in a crash event, thus being a potential game changer in the vehicle safety design process.