@article{Emami_2026,

title = {Robust-multi-task gradient boosting},

author = {Seyedsaman Emami and Gonzalo Martínez-Muñoz and Daniel Hernández-Lobato},

url = {http://dx.doi.org/10.1016/j.eswa.2025.130696},

doi = {10.1016/j.eswa.2025.130696},

issn = {0957-4174},

year  = {2026},

date = {2026-03-01},

journal = {Expert Systems with Applications},

volume = {303},

pages = {130696},

publisher = {Elsevier BV},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{11283058,

title = {On Noncoherent Multiple-Antenna Rayleigh Block-Fading Channels at Finite Blocklength},

author = {Chao Qi and Tobias Koch},

doi = {10.1109/TIT.2025.3641371},

year  = {2026},

date = {2026-01-01},

journal = {IEEE Transactions on Information Theory},

volume = {72},

number = {3},

pages = {1691-1724},

keywords = {Signal to noise ratio;Encoding;Dispersion;Receiving antennas;Transmitting antennas;Rayleigh channels;Lower bound;Error correction codes;Channel capacity;Accuracy;Channel dispersion;finite blocklength;high SNR;MIMO;normal approximation;Rayleigh fading;wireless communications},

pubstate = {published},

tppubtype = {article}

}

@article{BELENGUERLLORENS2026104070,

title = {Interpretable generative and discriminative learning for multimodal and incomplete clinical data},

author = {Albert Belenguer-Llorens and Carlos Sevilla-Salcedo and Janaina Mourão-Miranda and Vanessa Gómez-Verdejo},

url = {https://www.sciencedirect.com/science/article/pii/S1566253525011327},

doi = {https://doi.org/10.1016/j.inffus.2025.104070},

issn = {1566-2535},

year  = {2026},

date = {2026-01-01},

journal = {Information Fusion},

volume = {129},

pages = {104070},

abstract = {Real-world clinical problems are often characterized by multimodal data, usually associated with incomplete views and limited sample sizes in their cohorts, posing significant limitations for machine learning algorithms. In this work, we propose a Bayesian approach designed to efficiently handle these challenges while providing interpretable solutions. Our approach integrates (1) a generative formulation to capture cross-view relationships with a semi-supervised strategy, and (2) a discriminative task-oriented formulation to identify relevant information for specific downstream objectives. This dual generative-discriminative formulation offers both general understanding and task-specific insights; thus, it provides an automatic imputation of the missing views while enabling robust inference across different data sources. The potential of this approach becomes evident when applied to the multimodal clinical data, where our algorithm is able to capture and disentangle the complex interactions among biological, psychological, and sociodemographic modalities. Across nine multimodal datasets, OSIRIS improves the average AUC from about 0.87 for existing methods to 0.92 and increases balanced accuracy from 0.66 to 0.83, achieving state-of-the-art performance on all benchmarks. Notably, OSIRIS remains robust under missing-data conditions, consistently outperforming existing imputation-based approaches. These results show that explicitly disentangling generative and discriminative latent factors yields reliable multimodal learning in low-data regimes.},

keywords = {Bayesian, Discriminative, Generative, Incomplete, Latent},

pubstate = {published},

tppubtype = {article}

}

@article{López-Santiago_2026,

title = {Data-driven Informative Priors for Bayesian Inference with Quasiperiodic Data},

author = {Javier López-Santiago and Luca Martino and Joaquín Miguez and Gonzalo Vázquez-Vilar},

url = {https://doi.org/10.3847/1538-3881/ae2468},

doi = {10.3847/1538-3881/ae2468},

year  = {2026},

date = {2026-01-01},

journal = {The Astronomical Journal},

volume = {171},

number = {2},

pages = {61},

publisher = {The American Astronomical Society},

abstract = {Bayesian computational strategies for inference can be inefficient in approximating the posterior distribution in models that exhibit some form of periodicity. This is because the probability mass of the marginal posterior distribution of the parameter representing the period is usually highly concentrated in a very small region of the parameter space. Therefore, it is necessary to provide as much information as possible to the inference method through the parameter prior distribution. We intend to show that it is possible to construct a prior distribution from the data by fitting a Gaussian process (GP) with a periodic kernel. More specifically, we want to show that it is possible to approximate the marginal posterior distribution of the hyperparameter corresponding to the period in the kernel. Subsequently, this distribution can be used as a prior distribution for the inference method. We use an adaptive importance sampling method to approximate the posterior distribution of the hyperparameters of the GP. Then, we use the marginal posterior distribution of the hyperparameter related to the periodicity in order to construct a prior distribution for the period of the parametric model. This workflow is empirical Bayes, implemented as a modular (cut) transfer of a GP posterior for the period to the parametric model. We applied the proposed methodology to both synthetic and real data. We approximated the posterior distribution of the period of the GP kernel and then passed it forward as a posterior-as-prior with no feedback. Finally, we analyzed its impact on the marginal posterior distribution.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{LARESE2026,

title = {Modeling recurrent suicide attempts using probabilistic Hawkes processes},

author = {Darío C. Larese and Pablo M. Olmos and Antonio Artés-Rodríguez and Laura Arenas-Pijoan and Eugènia Nicolau-Subires and Vicent Llorca-Bofí and Enrique Baca-García and María Irigoyen-Otiñano and Jorge López-Castromán},

url = {https://www.sciencedirect.com/science/article/pii/S2950285326000116},

doi = {https://doi.org/10.1016/j.sjpmh.2026.01.003},

issn = {2950-2853},

year  = {2026},

date = {2026-01-01},

journal = {Spanish Journal of Psychiatry and Mental Health},

abstract = {Background 

Assessing the risk of suicide attempt recurrence requires integrating multiple clinical factors, including suicidal ideation and intent. Although clinical evaluation remains the most reliable method for estimating risk, few longitudinal mathematical models exist that leverage routine clinical data to predict recurrence dynamically. This gap limits the use of predictive analytics in suicide prevention. 

Methods 

We analyzed data from 1112 individuals from the MCOSUL Cohort (Lleida, Spain), who were treated for a suicide attempt, with a minimum 5-year follow-up or until death. Baseline sociodemographic and clinical variables were collected during structured assessment, and follow-up data were extracted from electronic health records. For each participant, Hawkes process parameters (μ, α, δ) were estimated using maximum likelihood and conditioned via a neural network. A Gaussian Mixture Model was applied to identify temporal risk profiles. 

Results 

Recurrence showed a temporal clustering pattern: 61.1% of repeat attempts occurred within 1 month of a previous event, and nearly all within 12 months. The model captured self-exciting dynamics and generated individualized survival and intensity curves. Five clusters emerged: a large low-risk heterogeneous group; a moderate-risk group; a predominantly male group with infrequent and less severe attempts; a high-risk group with multiple previous attempts; and a small but extreme group with severe and chronic recurrences. 

Conclusions 

Suicide attempt repetition in this cohort demonstrates self-exciting temporal behavior. Hawkes-based modeling enables dynamic, time-varying risk estimation and may offer advantages over traditional static prediction tools. Prospective validation should assess clinical integration, scalability, and utility for personalized suicide prevention.},

keywords = {Multiple suicide attempts, Personalized prediction, Probabilistic function, Risk assessment},

pubstate = {published},

tppubtype = {article}

}

@article{FERNANDEZSANCHEZ2026134400,

title = {Impact of evaluation noise in the context of multi-objective Bayesian optimization with the intrinsic coregionalization model},

author = {Daniel Fernández-Sánchez and Carlos Sevilla-Salcedo and Vanessa Gómez-Verdejo and Daniel Hernández-Lobato},

url = {https://www.sciencedirect.com/science/article/pii/S0925231226017984},

doi = {https://doi.org/10.1016/j.neucom.2026.134400},

issn = {0925-2312},

year  = {2026},

date = {2026-01-01},

journal = {Neurocomputing},

volume = {700},

pages = {134400},

abstract = {Multi-objective Bayesian optimization (MOBO) is a powerful framework for optimizing black-box functions with conflicting objectives and costly evaluations. A key component in MOBO is the probabilistic surrogate model used to guide the search for the optimum. A simple approach is to use independent Gaussian processes (GPs), one per objective. A more powerful method is to use multi-output GPs, such as those based on the intrinsic coregionalization model (ICM), to exploit correlations between objectives. However, with this model, often independent observation noise across objectives is assumed, which ignores part of the correlation structure. In this work, we isolate and analyze the role of observational noise in MOBO with ICM-based multi-output GPs. We consider models that can capture correlations in the observational noise, and study how the presence and structure of noise affect the joint predictive distribution, the considered acquisition function, and ultimately the optimization performance.Through theoretical analysis and extensive synthetic and real-world experiments, we show that multi-output GPs can significantly improve optimization efficiency when the evaluations are noisy. Importantly, however, when the evaluations are noiseless, we observe, both theoretically and empirically, only small gains by considering multi-output GPs, and cheaper independent GPs may be preferred. Furthermore, our synthetic experiments show that explicitly modeling noise correlations provides additional advantages when the objectives themselves are negatively correlated. In real-world problems, however, we have observed mixed results regarding modeling noise correlations. There are small gains in an ensemble benchmark and clearer gains, similar to those observed in the synthetic experiments, in a neural-network benchmark. These findings provide practical insights into when and how to use ICMs for MOBO.},

keywords = {Bayesian optimization, Multi-output Gaussian process, Multiple objectives},

pubstate = {published},

tppubtype = {article}

}

@article{11556476,

title = {Fundamental Limits of Noncoherent Massive Random Access Networks},

author = {Grace Villacrés and Tobias Koch and Gonzalo Vazquez-Vilar},

doi = {10.1109/TIT.2026.3701621},

year  = {2026},

date = {2026-01-01},

journal = {IEEE Transactions on Information Theory},

pages = {1-1},

keywords = {Cells (biology);Interference;Modeling;Upper bound;Sequential analysis;Limiting;Lower bound;Random variables;Mutual information;Entropy;Channel capacity;fading channel;massive multiple access;noncoherent;random access;wireless networks},

pubstate = {published},

tppubtype = {article}

}

@article{Blázquez-Sánchez2025,

title = {Automated web-based typing of Clostridioides difficile ribotypes via MALDI-TOF MS},

author = {Mario Blázquez-Sánchez and Alejandro Guerrero-López and Ana Candela and Albert Belenguer-Llorens and José Miguel Moreno and Carlos Sevilla-Salcedo and María Sánchez-Cueto and Manuel J. Arroyo and Mark Gutiérrez-Pareja and Vanessa Gómez-Verdejo and Pablo M. Olmos and Luis Mancera and Patricia Muñoz and Mercedes Marín and Luis Alcalá and David Rodríguez-Temporal and Belén Rodríguez-Sánchez},

doi = {10.1186/s12859-025-06200-6},

issn = {1471-2105},

year  = {2025},

date = {2025-06-19},

journal = {BMC Bioinformatics},

volume = {26},

number = {1},

publisher = {Springer Science and Business Media LLC},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{info:doi/10.2196/63962,

title = {Emotion Forecasting: A Transformer-Based Approach},

author = {Leire Paz-Arbaizar and Jorge Lopez-Castroman and Antonio Artés-Rodríguez and Pablo M Olmos and David Ramírez},

url = {https://doi.org/10.2196/63962},

doi = {10.2196/63962},

issn = {1438-8871},

year  = {2025},

date = {2025-03-18},

journal = {J Med Internet Res},

volume = {27},

pages = {e63962},

abstract = {Background: Monitoring the emotional states of patients with psychiatric problems has always been challenging due to the noncontinuous nature of clinical assessments, the effect of the health care environment, and the inherent subjectivity of evaluation instruments. However, mental states in psychiatric disorders exhibit substantial variability over time, making real-time monitoring crucial for preventing risky situations and ensuring appropriate treatment. Objective: This study aimed to leverage new technologies and deep learning techniques to enable more objective, real-time monitoring of patients. This was achieved by passively monitoring variables such as step count, patient location, and sleep patterns using mobile devices. We aimed to predict patient self-reports and detect sudden variations in their emotional valence, identifying situations that may require clinical intervention. Methods: Data for this project were collected using the Evidence-Based Behavior (eB2) app, which records both passive and self-reported variables daily. Passive data refer to behavioral information gathered via the eB2 app through sensors embedded in mobile devices and wearables. These data were obtained from studies conducted in collaboration with hospitals and clinics that used eB2. We used hidden Markov models (HMMs) to address missing data and transformer deep neural networks for time-series forecasting. Finally, classification algorithms were applied to predict several variables, including emotional state and responses to the Patient Health Questionnaire-9. Results: Through real-time patient monitoring, we demonstrated the ability to accurately predict patients' emotional states and anticipate changes over time. Specifically, our approach achieved high accuracy (0.93) and a receiver operating characteristic (ROC) area under the curve (AUC) of 0.98 for emotional valence classification. For predicting emotional state changes 1 day in advance, we obtained an ROC AUC of 0.87. Furthermore, we demonstrated the feasibility of forecasting responses to the Patient Health Questionnaire-9, with particularly strong performance for certain questions. For example, in question 9, related to suicidal ideation, our model achieved an accuracy of 0.9 and an ROC AUC of 0.77 for predicting the next day's response. Moreover, we illustrated the enhanced stability of multivariate time-series forecasting when HMM preprocessing was combined with a transformer model, as opposed to other time-series forecasting methods, such as recurrent neural networks or long short-term memory cells. Conclusions: The stability of multivariate time-series forecasting improved when HMM preprocessing was combined with a transformer model, as opposed to other time-series forecasting methods (eg, recurrent neural network and long short-term memory), leveraging the attention mechanisms to capture longer time dependencies and gain interpretability. We showed the potential to assess the emotional state of a patient and the scores of psychiatric questionnaires from passive variables in advance. This allows real-time monitoring of patients and hence better risk detection and treatment adjustment.},

keywords = {affect; emotional valence; machine learning; mental disorder; monitoring; mood; passive data; Patient Health Questionnaire-9; PHQ-9; psychological distress; time-series forecasting},

pubstate = {published},

tppubtype = {article}

}

@article{CONTINELLI2025105155,

title = {Combined model-based and data-driven approach for the control of a soft robotic neck},

author = {Nicole A. Continelli and Luis F. Nagua and Pablo M. Olmos and Concepción A. Monje},

url = {https://www.sciencedirect.com/science/article/pii/S0921889025002520},

doi = {https://doi.org/10.1016/j.robot.2025.105155},

issn = {0921-8890},

year  = {2025},

date = {2025-01-01},

journal = {Robotics and Autonomous Systems},

volume = {194},

pages = {105155},

abstract = {This paper delves into the potential of integrating model-based and data-driven techniques for controlling the performance of a soft robotic neck. Artificial intelligence (AI) methods, such as machine learning and deep learning, have shown their applicability in modelling and controlling robotic systems with complex nonlinear behaviours. However, model-based approaches have also proven to be effective analytical alternatives, even if they rely on simplified approximations of the robot model. The control system proposed in this work combines the closed loop analytical model of the soft robotic neck with a Multi-Layer Perceptron (MLP) network trained to minimise the neck pose error. The MLP undergoes training with three different data treatments, and the results are compared to determine the most effective one. The experimental results obtained demonstrate the robustness of the proposed technique and its potential as an alternative to classical solutions, whether purely based on analytical models or data-driven models.},

keywords = {Data processing, Machine learning, Model-based control, Multi-layer perceptron, Neural network, Soft robotics},

pubstate = {published},

tppubtype = {article}

}

@article{doi:10.1200/JCO.2025.43.16_suppl.1651,

title = {Remote physiologic and behavioral monitoring to predict early treatment response in metastatic cancer: High-Definition Oncology study (HDOs) preliminary results.},

author = {Leire Paz-Arbaizar and María Sauras and Sonia Pernas and David Vicente and Rosario Garcia-Campelo and Josefa Terrasa and Ramon Colomer Bosch and Ruth Vera and Desirée Jiménez and Santiago Gonzalez- Santiago and Begona Bermejo and Antonio López-Alonso and Berta Nasarre and Leonardo Garma and Pablo Martínez Olmos and Antonio Artés Rodríguez and Miguel Quintela-Fandino},

url = {https://ascopubs.org/doi/abs/10.1200/JCO.2025.43.16_suppl.1651},

doi = {10.1200/JCO.2025.43.16_suppl.1651},

year  = {2025},

date = {2025-01-01},

journal = {Journal of Clinical Oncology},

volume = {43},

number = {16_suppl},

pages = {1651-1651},

abstract = { 1651Background: Emerging evidence suggests that behavioral, physiologic or emotional factors may act as real-time indicators of treatment response, with potential as modifiable factors. Advances in remote monitoring technologies provide passive (e.g., heart rate, sleep patterns) and active (e.g., self-reported emotions) data. HDOs collects such data and serial -omics from 300 women with metastatic cancer to identify novel markers, understand disease trajectories and develop a digital twin for individualized care. We present data from 25% accrual. Methods: Women receiving first-line treatment for metastatic colorectal, lung, or hormone-positive breast cancer were eligible. Patients continuously wore a smartwatch and used the EB2 App to capture step count (SC), sleep duration (SL), phone usage (PU), time at home (TH), location clusters (LC), mean (MHR) and minimum (mHR) heart rate, mean (MSHR) and minimum (mSHR) sleeping heart rate and sleeping oxygen saturation (SOS). Emotional valence was self-reported from a list of 20 emotions and classified as negative (-1), neutral (0) or positive (+1). Aim 1: to explore the relationship between the variables and response (CB: CR+PR+SD vs. PD) at the first CT scan at day +90 analyzing data from days 1-15 and 75-90 (Mann-Whitney U). Aim 2: to find Response-Associated Behavioral Patterns (RABPs) associated with CB or PD. First, Daily Behavioral Profiles (DBPs) are obtained using unsupervised learning models from > 2 million days of smartwatch and App data (external set). After identifying 256 DBPs with the VQ-VAE model, Latent Dirichlet Allocation defined RABPS based on the frequency and abundance of DBPs per patient. RABPs were compared among classes (response type, age group) using X2. Bilateral P values < 0.01 were deemed significant. Results: from May 2023 to April 2024, 77 female patients (median age 61; 28-80) were accrued (46 Breast, 23 Lung, 8 Colorectal). At first CT, 72 (93.5%) achieved CB while 5 (6.5%) had PD. During days 1-15, CB patients showed lower PU (2.4 vs. 3.9 hours; P = 0.002), TH (18.5 vs. 22 hours; P = 2* 10^-7), MHR (78 vs. 88 bpm), mHR (59 vs 70 bpm), MSHR (75 vs 88 bpm) mSHR (66 vs 78 bpm) (all Ps < 10^-10) and reported more negative EV. The trends persisted in days 76-90 in addition to SC (7235 vs 4038 steps/day; P = 0.00001) and decreased SOS (90.1% vs. 92.6%; P = 1.5*10^-8). Six RABPS were identified. Patients < 60 yo displayed more often RABPs 1, 2 and 5 (84% vs. 16%; P = 0.02). RABP1 breast cancer and RABP5 lung cancer patients were more likely to experience PD vs. CB (75% vs. 24%; P = 0.08; and 69% vs.19%},

note = {PMID:},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{10.1093/mnras/staf2256,

title = {A recurrent 70–100 minute quasi-periodic pulsation in the intermediate-aged mid-M dwarf GJ 3512},

author = {J López-Santiago and F Reale and G Micela and L Martino and G Vázquez-Vilar and J Miguez},

url = {https://doi.org/10.1093/mnras/staf2256},

doi = {10.1093/mnras/staf2256},

issn = {0035-8711},

year  = {2025},

date = {2025-01-01},

journal = {Monthly Notices of the Royal Astronomical Society},

pages = {staf2256},

abstract = {We report the discovery of a recurrent quasi-periodic pulsation (QPP) in the late-M dwarf GJ 3512 (M5.5V) using multiple TESS datasets. A strong signal with a period of 70–100 minutes was detected in wavelet analyses of the two-minute cadence light curve from Sector 20. This signal was detected also in observations from Sectors 47 and 60. The QPP persisted for weeks in sector 20 and spanned nearly three years of TESS coverage. There was no significant damping between major flares. This behavior contrasts with that of previously reported stellar QPPs, which are confined to individual flares and decay on timescales of minutes to hours. The oscillation amplitude is at the milli-magnitude level. A pulsation origin is discarded since theoretical instability strips for 100-minute pulsations are restricted to pre-main sequence stars, while GJ 3512 is an intermediate age (2-8 Gyr) main-sequence dwarf. The persistence across independent TESS sectors discards an instrumental artifact origin and points to a likely coronal origin instead, such as oscillatory reconnection or thermal non-equilibrium cycles in large active regions. This represents the first detection of a likely sustained QPP with these characteristics in a late-type star, highlighting the need for further investigation into physical mechanisms behind such variability.},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{BELENGUERLLORENS2025111887,

title = {Unified Bayesian representation for high-dimensional multi-modal biomedical data for small-sample classification},

author = {Albert Belenguer-Llorens and Carlos Sevilla-Salcedo and Jussi Tohka and Vanessa Gómez-Verdejo},

url = {https://www.sciencedirect.com/science/article/pii/S0952197625018895},

doi = {https://doi.org/10.1016/j.engappai.2025.111887},

issn = {0952-1976},

year  = {2025},

date = {2025-01-01},

journal = {Engineering Applications of Artificial Intelligence},

volume = {160},

pages = {111887},

abstract = {The increasing availability of multi-modal medical data, including neuroimaging, genetic profiles, and clinical measurements, offers unprecedented opportunities for advancing disease diagnosis and prognosis. However, integrating these heterogeneous data sources poses significant challenges due to their high dimensionality, redundancy, and small sample sizes, which hinder the effectiveness of traditional machine learning models. To overcome these challenges, we present the BAyesian Latent Data Unified Representation model (BALDUR), a novel Bayesian algorithm designed to deal with multi-modal datasets and small sample sizes in high-dimensional settings while providing explainable solutions. To do so, the proposed model combines within a common latent space the different data views to extract the relevant information to solve the classification task and prune out the irrelevant/redundant features/data views. Furthermore, to provide generalizable solutions in small sample size scenarios, BALDUR efficiently integrates dual kernels over the views with a small sample-to-feature ratio. Finally, its linear nature ensures the explainability of the model outcomes, allowing its use for biomarker identification. This model was tested over two different neurodegeneration datasets, outperforming the state-of-the-art models and detecting features aligned with markers already described in the scientific literature.},

keywords = {Bayesian modeling, Machine learning health applications, Multi-modal data, Wide-data},

pubstate = {published},

tppubtype = {article}

}

@article{BELENGUERLLORENS2025110985,

title = {Addressing wide-data studies of gene expression microarrays with the Relevance Feature and Vector Machine},

author = {Albert Belenguer-Llorens and Carlos Sevilla-Salcedo and Emilio Parrado-Hernández and Vanessa Gómez-Verdejo},

url = {https://www.sciencedirect.com/science/article/pii/S001048252501337X},

doi = {https://doi.org/10.1016/j.compbiomed.2025.110985},

issn = {0010-4825},

year  = {2025},

date = {2025-01-01},

journal = {Computers in Biology and Medicine},

volume = {197},

pages = {110985},

abstract = {This paper presents the Relevance Feature and Vector Machine (RFVM), a novel Bayesian model that addresses the wide-data challenges of gene expression microarrays, ensuring interpretability in both feature and sample spaces. The wide-data problem occurs when Machine Learning algorithms encounter databases with significantly more features than observations, as commonly seen in gene expression microarrays. To address these challenges, RFVM operates in the dual space, enabling effective parameter inference with small patient cohorts, thereby avoiding overfitting and ensuring generalizable solutions. The core innovation of RFVM lies in its two-way sparsity approach, incorporating priors over both primal and dual variables to perform joint feature and sample selection. As we will show, this capability is critical in wide-data clinical settings, as identifying relevant patients enhances the feature selection process, while effective feature selection, in turn, improves sample identification. This interaction results in more compact solutions and boosts the interpretability and performance of the final model. The RFVM’s capabilities are validated against several models on multiple gene expression microarray datasets. Results demonstrate that RFVM achieves superior performance in diagnostic tasks while producing the most compact solutions. Furthermore, the selected genes align with known biomarkers from the medical literature, highlighting the potential of the model as a clinical tool.},

keywords = {Feature selection, Gene expression microarrays, Low-sample-to-feature ratio, Probabilistic machine learning, Sample selection},

pubstate = {published},

tppubtype = {article}

}

@article{hendrych2025secant,

title = {Secant Line Search for Frank-Wolfe Algorithms},

author = {Deborah Hendrych and Sebastian Pokutta and Mathieu Besançon and David Martínez-Rubio},

url = {https://arxiv.org/abs/2501.18775},

year  = {2025},

date = {2025-01-01},

journal = {arXiv preprint arXiv:2501.18775},

keywords = {},

pubstate = {published},

tppubtype = {article}

}

@article{FERNANDEZSANCHEZ2025131674,

title = {Joint entropy search for multi-objective Bayesian optimization with constraints and multiple fidelities},

author = {Daniel Fernández-Sánchez and Daniel Hernández-Lobato},

url = {https://www.sciencedirect.com/science/article/pii/S092523122502346X},

doi = {https://doi.org/10.1016/j.neucom.2025.131674},

issn = {0925-2312},

year  = {2025},

date = {2025-01-01},

journal = {Neurocomputing},

volume = {657},

pages = {131674},

abstract = {Bayesian optimization (BO) methods can be used to solve efficiently problems with several objectives and constraints. Each objective and constraint is considered a black-box function that is expensive to evaluate, lacking a closed-form expression. BO methods use a model of each black-box to guide the search for the problem’s solution. Specifically, they make intelligent decisions about where each black-box function should be evaluated next with the goal of finding the solution using a few evaluations only. Sometimes, however, the black-boxes may be evaluated at different fidelity levels. A lower fidelity is simply a cheap proxy for the corresponding black-box. These lower fidelities correlate with the actual black-boxes to optimize and can, therefore, be used to reduce the overall cost of solving the optimization problem. Here, we propose Multi-fidelity Joint Entropy Search for Multi-objective Bayesian Optimization with Constraints (MF-JESMOC), a BO method for solving the aforementioned problems. MF-JESMOC chooses the next point, and fidelity level at which to evaluate the black-boxes, as the combination that is expected to reduce the most the joint entropy of the Pareto set and the Pareto front, normalized by the fidelity’s evaluation cost. We use Deep Gaussian processes to model each black-box and the dependencies between fidelities. These are powerful probabilistic models that can learn the dependency structure among fidelity levels of each black-box. Several experiments show that MF-JESMOC outperforms other state-of-the-art methods for multi-objective BO with constraints and different fidelity levels in both synthetic and real-world problems.},

keywords = {Bayesian optimization, Constrained multi-objective optimization, Information theory, Multi-fidelity optimization},

pubstate = {published},

tppubtype = {article}

}

@article{FERNANDEZSANCHEZ2025113612,

title = {Alpha entropy search for new information-based Bayesian optimization},

author = {Daniel Fernández-Sánchez and Eduardo C. Garrido-Merchán and Daniel Hernández-Lobato},

url = {https://www.sciencedirect.com/science/article/pii/S0950705125006586},

doi = {https://doi.org/10.1016/j.knosys.2025.113612},

issn = {0950-7051},

year  = {2025},

date = {2025-01-01},

journal = {Knowledge-Based Systems},

volume = {322},

pages = {113612},

abstract = {Bayesian optimization (BO) methods based on information theory have obtained state-of-the-art results in several tasks. These techniques rely on the Kullback–Leibler (KL) divergence to compute the acquisition function. We introduce a novel information-based class of acquisition functions for BO called Alpha Entropy Search (AES). AES is based on the alpha-divergence, which generalizes the KL-divergence. Iteratively, AES selects the next evaluation point as the one whose associated target value has the highest level of dependency with respect to the location and associated value of the global maximum of the optimization problem. Dependency is measured in terms of the alpha-divergence, as an alternative to the KL-divergence. Intuitively, this favors evaluating the objective function at the most informative points about the global maximum. The alpha-divergence has a free parameter α, which determines the behavior of the divergence, balancing local and global differences. Therefore, different values of α result in different acquisition functions. AES acquisition lacks a closed-form expression. However, we propose an efficient and accurate approximation using a truncated Gaussian distribution. In practice, the value of α can be chosen by the practitioner, but here we suggest using a combination of acquisition functions obtained by simultaneously considering a range of α values. We provide an implementation of AES in BOTorch and we evaluate its performance in synthetic, benchmark, and real-world experiments involving the tuning of the hyper-parameters of a deep neural network. These experiments show that AES performance is competitive with other information-based acquisition functions such as JES, MES, or PES.},

keywords = {Alpha-divergence, Bayesian optimization, Entropy search, Information theory},

pubstate = {published},

tppubtype = {article}

}