Skip to main content

Resultados relacionados

Resultados totales (Incluyendo duplicados): 4
Encontrada(s) 1 página(s)

Repositori Institucional UIB
Publication Capítulo o parte de un libro (BookPart)

Aggregating distances with uncertainty: the modular (pseudo-)metric case

  • Bibiloni-Femenias, M. D. M.
  • Miñana, J. J.
  • Valero, O.
[eng] Many real problems involve the use of metrics to be solved. In such problems to merge some measurements obtained through the used metric is often needed in order to make a working decision. So functions that merge a collection of metrics providing a new one as output have roused interest of different authors. Nevertheless, the axiomatic that defines the concept of metric limits their applicability in some situations. Motivated by this fact, different generalizations of the concept of metric can be found in the literature. Among them, it is worth mentioning the notion of modular metric introduced by Chistyakov. This kind of metric involves uncertainty due to the inclusion of a parameter in the distance evaluation. The aim of this paper consists in characterizing those functions that merge a collection of modular (pseudo-)metrics into a new modular (pseudo-)metric, all of them defined on the same set.
Repositori Institucional UIB
Publication Capítulo o parte de un libro (BookPart)

On the Use of Modular Indistinguishability Operators in RBFNN-like Models

  • Ortiz, A.
  • Valero, O.
  • Miñana, J.J.
[eng] Radial Basis Function Neural Networks (RBFNN) have become popular machine learning models with a simple structure but at the same time strong non-linear function approximation and effective modeling capabilities. In this work, we explore the use of Modular Indistinguishability Operators (MIO) in RBFNN-like structures to replace the RBFs that populate the hidden layer, to give rise to MIO-based Neural Networks (MIO-NN). In this respect, we introduce a new distance function and prove that it is a modular metric, to next use it to derive two MIOs to be evaluated as the key component of MIO-NNs. As an additional contribution, we describe Self-Defining MIO-NN (SD-MIO-NN) as an approach capable of configuring MIO-NNs in a parameterless way. SD-MIO-NN comprises a first step that defines the size of the hidden layer, a second step that determines the parameters of the hidden neurons and a last step that calculates the weights of the hidden-to-output layer connections. The experimental results show the effectiveness of the proposed MIOs for multi-class classification, and by extension of SDMIO-NN, which in turn compares well with other similar solutions.
Repositori Institucional UIB
Publication Artículo científico (JournalArticle)

From flight to insight: An end-to-end aerial inspection framework for 3D modeling and defect detection of infrastructures

  • Bonnin-Pascual, Francisco
  • Garcia-Fidalgo, Emilio
  • Tauler-Rossello, Antoni
  • Ortiz, Alberto
[eng] This work presents a complete inspection framework based on Unmanned Aerial Vehicles (UAVs) specifically designed for maritime environments, integrating autonomous data acquisition, high-fidelity three-dimensional (3D) reconstruction, and Artificial Intelligence (AI)-based defect detection and localization. The proposed system combines a custom aerial robotic platform equipped with onboard Light Detection And Ranging (LiDAR) and imaging sensors, a hybrid pose estimation approach that merges real-time LiDAR odometry with an offline refinement stage, and a robust processing pipeline that produces metrically accurate 3D models of vessel structures. Structural anomalies identified through deep learning-based image analysis are projected directly onto the reconstructed models, providing an intuitive spatial representation of the detected defects. The framework was experimentally validated through field trials conducted aboard real maritime vessels within the scope of the BugWright2 project. The results demonstrate accurate reconstruction and reliable defect localization even in environments without satellite-based positioning availability.
Repositori Institucional UIB
Publication Capítulo o parte de un libro (BookPart)

Addressing Multi-class Classification Tasks by means of RBFNN-like Models using Modular Indistinguishability Operators

  • Ortiz, A.
  • Valero, O.
[eng] As a machine learning model, a Radial Basis Function Neural Network (RBFNN) is an artificial neural network that comprises a single hidden layer whose neurons implement a set of radial-basis functions that are linearly combined at the output layer. The most popular form of RBFNN makes use of the Gaussian RBF combined with the Euclidean distance. In this work, we explore the use of Modular Indistinguishability Operators (MIO) in the hidden layer of RBFNN-like structures in replacement of Gaussian RBFs. Moreover, we introduce a new modular metric m(t, p, q) and next use it to derive a specific MIO IL(t, a, b) from the Luckasiewicz t-norm. The resulting MIO is next used as the activation function of the hidden neurons, whose parameters are determined on the basis of the samples of the underlying dataset. The performance of the new machine learning model is evaluated through a set of publicly available datasets in the context of multi-class classification tasks.