This paper introduces a novel approach to learn multi-task regression models with constrained architecture complexity. The proposed model, named RFF-BLR, consists of randomised feedforward neural network two fundamental characteristics: single hidden layer whose units implement the random Fourier features that approximate an RBF kernel, and Bayesian formulation optimises weights connecting output layers. RFF-based inherits robustness kernel methods. enables promoting multioutput sparsity:...

Feature selection in noisy label scenarios remains an understudied topic. We propose a novel genetic algorithm-based approach, the Noise-Aware Multi-Objective Selection Genetic Algorithm (NMFS-GA), for selecting optimal feature subsets binary classification with labels. NMFS-GA offers unified framework that are both accurate and interpretable. evaluate on synthetic datasets noise, Breast Cancer dataset enriched features, real-world ADNI dementia conversion prediction. Our results indicate...

This paper presents the Relevance Feature and Vector Machine (RFVM), a novel model that addresses challenges of fat-data problem when dealing with clinical prospective studies. The refers to limitations Learning (ML) algorithms working databases in which number features is much larger than samples (a common scenario certain medical fields). To overcome such limitations, RFVM incorporates different characteristics: (1) A Bayesian formulation enables infer its parameters without overfitting...

Abstract The implementation of Matrix-assisted laser desorption ionization–time flight (MALDI-TOF) mass spectrometry has had a profound impact on clinical microbiology, facilitating rapid bacterial identification through protein profile analysis. However, the application this technique is limited by challenges related to reproducibility and variability spectra, particularly in distinguishing closely strains, as exemplified typification Clostridioides difficile ribotypes. This thesis...

Healthcare-associated infections (HAIs) are a significant concern within hospital environments, with the World Health Organization (WHO) identifying them as major source of bacteriological infections. HAIs affect millions patients annually, leading to substantial morbidity and mortality. However, proportion preventable through early detection appropriate intervention isolation. Traditional methods for bacterial species strains, such antigen tests, often time-consuming hamper realtime...

We present 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 common latent space different data views extract relevant information solve classification task prune out irrelevant/redundant features/data views. Furthermore, provide generalizable solutions size scenarios, BALDUR efficiently integrates dual kernels over...

In this paper, we propose a novel Machine Learning Model based on Bayesian Linear Regression intended to deal with the low sample-to-variable ratio typically found in neuroimaging studies and focusing mental disorders. The proposed model combines feature selection capabilities formulation dual space which, turn, enables efficient work data. Thus, have tested algorithm real MRI data from an animal of schizophrenia. results show that our proposal efficiently predicts diagnosis and, at same...

Abstract Matrix-Assisted Laser Desorption Ionization Time-Of-Flight (MALDI-TOF) Mass Spectrometry (MS) is a reference method for microbial identification and it can be used to predict Antibiotic Resistance (AR) when combined with artificial intelligence methods. However, current solutions need time-costly preprocessing steps, are difficult reproduce due hyperparameter tuning, hardly interpretable, do not pay attention epidemiological differences inherent data coming from different centres,...

Multi-view datasets offer diverse forms of data that can enhance prediction models by providing complementary information. However, the use multi-view leads to an increase in high-dimensional data, which poses significant challenges for lead poor generalization. Therefore, relevant feature selection from is important as it not only addresses generalization but also enhances interpretability models. Despite success traditional methods, they have limitations leveraging intrinsic information...

Abstract A fundamental problem of supervised learning algorithms for brain imaging applications is that the number features far exceeds subjects. In this paper, we propose a combined feature selection and extraction approach multiclass problems. This method starts with bagging procedure which calculates sign consistency multivariate analysis (MVA) projection matrix feature-wise to determine relevance each feature. measure provides parsimonious matrix, hypothesis test automatically selected...

Multi-view problems can be faced with latent variable models since they are able to find low-dimensional projections that fairly capture the correlations among multiple views characterise each datum. On other hand, high-dimensionality and non-linear issues traditionally handled by kernel methods, inducing a (non)-linear function between projection data itself. However, usually come scalability exposition overfitting. Here, we propose merging both approaches into single model so exploit best...

The Bayesian approach to feature extraction, known as factor analysis (FA), has been widely studied in machine learning obtain a latent representation of the data. An adequate selection probabilities and priors these bayesian models allows model better adapt data nature (i.e. heterogeneity, sparsity), obtaining more representative space. objective this article is propose general FA framework capable modelling any problem. To do so, we start from Inter-Battery Factor Analysis (BIBFA) model,...

Machine learning techniques typically applied to dementia forecasting lack in their capabilities jointly learn several tasks, handle time dependent heterogeneous data and missing values. In this paper, we propose a framework using the recently presented SSHIBA model for different tasks on longitudinal with The method uses Bayesian variational inference impute values combine information of views. This way, can data-views from time-points common latent space relations between each time-point...

Real-world databases are complex and usually require dealing with heterogeneous mixed data types making the exploitation of shared information between views a critical issue. For this purpose, recent studies based on deep generative models merge all into nonlinear latent space, which can share among views. However, solution limits model's interpretability, flexibility, modularity. We propose novel method to overcome these limitations by combining multiple Variational AutoEncoders (VAE)...

This paper introduces a novel approach for multi-task regression that connects Kernel Machines (KMs) and Extreme Learning (ELMs) through the exploitation of Random Fourier Features (RFFs) approximation RBF kernel. In this sense, one contributions shows proposed models, KM ELM formulations can be regarded as two sides same coin. These termed RFF-BLR, stand on Bayesian framework simultaneously addresses main design goals. On hand, it fits multitask regressors based KMs endowed with kernels....