L'objectiu principal del grup és avançar en les metodologies que conformen l'àrea de la infor-màtica tova (Soft Computing), així com investigar possibles hibridacions d'aquestes amb l'ob-jectiu de millorar-ne el rendiment i la fiabilitat. També és un objectiu important aplicar les metodologies desenvolupades a problemes reals en àrees com la medicina, ecologia, farmacoprote-òmica, e-learning, etc. El grup ha publicat més de 150 articles de revista, més de 300 articles en congressos, 50 capítols de llibres i 60 reports de recerca.

http://futur.upc.edu/SOCO

The term "soft computing" was coined in the nineties, and it describes the combined use of a variety of computational approaches that have been developed over the last few decades, which include but are not limited to fuzzy systems, neural networks and evolutionary algorithms.

Despite their obvious differences, a common trait of these fields is the abandonment of binary logic, static analytical models, rigid classifications and deterministic searches. In an ideal problem description, the systems that require modelling and/or control would be defined completely and precisely. In this case, formal reasoning systems can be used to associate Boolean values to state descriptions or physical systems' behaviour. Nevertheless, when tackling real-world problems, it is not unusual to find them incompletely or badly defined, which makes it difficult to model them and requires large search spaces. As a result, precise models, should they exist, might turn out to be impractical and/or costly. Usually, the relevant information that is available is presented either as empirical, a priori knowledge or as input-output instance descriptions of the systems' behaviour. This makes the use of approximate reasoning systems necessary, as they can flexibly cope with such far-from-perfect information.

The main goal of the SOCO group is to make progress in the state-of-the-art development of soft-computing methodologies, as well as to research their possible hybridisation in order to improve their robustness and efficacy. The SOCO group is currently working on the following research lines:

* Feature selection and dimensionality reduction

* Fuzzy systems (Fuzzy Inductive Reasoning, FIR)

* Artificial neural networks (feed-forward, recurrent, heterogeneous)

* Unsupervised probabilistic models

* Genetic algorithms and evolutionary strategies

* Pattern recognition and computer vision

* Hybrid soft-computing methods, including the following:

- Neural networks and support vector machines

- Fuzzy Inductive Reasoning and simulated annealing

- Fuzzy Inductive Reasoning and genetic algorithms

- Frequency selection for neural networks

- Cooperation of local experts for inductive reasoning

- Incremental construction of hybrid recurrent neural networks

The application of these methodologies to real-world problems is also one of the group's goals. The group has carried out research in the following areas of application:

* Medical (human central nervous system, cancer prediction, diagnosis, cognitive neuroscience, etc.)

* Biological (growth of white shrimp)

* Ecological (analysis of pollutant concentration in urban areas and ecological status modelling of streams)

http://futur.upc.edu/SOCO

The term "soft computing" was coined in the nineties, and it describes the combined use of a variety of computational approaches that have been developed over the last few decades, which include but are not limited to fuzzy systems, neural networks and evolutionary algorithms.

Despite their obvious differences, a common trait of these fields is the abandonment of binary logic, static analytical models, rigid classifications and deterministic searches. In an ideal problem description, the systems that require modelling and/or control would be defined completely and precisely. In this case, formal reasoning systems can be used to associate Boolean values to state descriptions or physical systems' behaviour. Nevertheless, when tackling real-world problems, it is not unusual to find them incompletely or badly defined, which makes it difficult to model them and requires large search spaces. As a result, precise models, should they exist, might turn out to be impractical and/or costly. Usually, the relevant information that is available is presented either as empirical, a priori knowledge or as input-output instance descriptions of the systems' behaviour. This makes the use of approximate reasoning systems necessary, as they can flexibly cope with such far-from-perfect information.

The main goal of the SOCO group is to make progress in the state-of-the-art development of soft-computing methodologies, as well as to research their possible hybridisation in order to improve their robustness and efficacy. The SOCO group is currently working on the following research lines:

* Feature selection and dimensionality reduction

* Fuzzy systems (Fuzzy Inductive Reasoning, FIR)

* Artificial neural networks (feed-forward, recurrent, heterogeneous)

* Unsupervised probabilistic models

* Genetic algorithms and evolutionary strategies

* Pattern recognition and computer vision

* Hybrid soft-computing methods, including the following:

- Neural networks and support vector machines

- Fuzzy Inductive Reasoning and simulated annealing

- Fuzzy Inductive Reasoning and genetic algorithms

- Frequency selection for neural networks

- Cooperation of local experts for inductive reasoning

- Incremental construction of hybrid recurrent neural networks

The application of these methodologies to real-world problems is also one of the group's goals. The group has carried out research in the following areas of application:

* Medical (human central nervous system, cancer prediction, diagnosis, cognitive neuroscience, etc.)

* Biological (growth of white shrimp)

* Ecological (analysis of pollutant concentration in urban areas and ecological status modelling of streams)

http://futur.upc.edu/SOCO

Enviaments recents

  • Generative topographic mapping as a constrained mixture of student t-distributions: theoretical developments 

    Vellido Alcacena, Alfredo (2004-09)
    Report de recerca
    Accés obert
    The Generative Topographic Mapping (GTM: Bishop et al. 1998a), a non-linear latent variable model, was originally defined as constrained mixture of Gaussians. Gaussian mixture models are known to lack robustness in the ...
  • Maximizing the margin with feed-forward neural networks 

    Romero Merino, Enrique (2001-07)
    Report de recerca
    Accés obert
    Feed-forward Neural Networks (FNNs) and Support Vector Machines (SVMs) are two machine learning frameworks developed from very different starting points of view. The solutions obtained by the respective frameworks may ...
  • Predicción a largo plazo de la concentración de ozono usando la metodología de razonamiento inductivo difuso 

    Gómez Miranda, Pilar; Nebot Castells, M. Àngela; Múgica Álvarez, Francisco (2001-09)
    Report de recerca
    Accés obert
    En este reporte se ha realizado un primer estudio para conocer la capacidad de la metodología de Razonamiento Inductivo Difuso (FIR), en la identificación de modelos para la predicción a largo plazo de las concentraciones ...
  • Function aproximation with SAOCIF: a general sequential method and a particular algorithm with feed-forward neural networks 

    Romero Merino, Enrique (2001-10)
    Report de recerca
    Accés obert
    A sequential method for approximating vectors in Hilbert spaces, called Sequential Approximation with Optimal Coefficients and Interacting Frequencies (SAOCIF), is presented. SAOCIF combines two key ideas. The first one ...
  • Algorithmes d'entraînement local de RBF 

    Quartier, Benoit; Belanche Muñoz, Luis Antonio (2001-10)
    Report de recerca
    Accés obert
    The aim of this work is to study the effect of locality in classification tasks with radial basis function neural networks (RBFNN). The networks are trained in a three stage process. Firstly, the data are decomposed ...
  • Comparison of Methods to Predict Ozone Concentration 

    Orozco Luquero, Jorge (2004-01)
    Report de recerca
    Accés obert
    Several methods have been applied to the prediction of ozone concentration. In this work, an Heterogeneous Neural Network (HNN) is used to perform the same task. Different capabilities of HNN are exploited like imprecision ...
  • Detecció i identificació de falles en una xarxa de distribució d'aigües 

    Escobet Canal, Antoni; Nebot Castells, M. Àngela (2001-04)
    Report de recerca
    Accés obert
    This technical report deals with two of the main tasks of Fault Monitoring Systems (FMS): fault detection and fault identification. During fault detection, the FMS should recognize that the plant behavior is abnormal, ...
  • A machine learning pipeline for supporting differentiation of glioblastomas from single brain metastases 

    Mocioiu, Victor; de Barros, Nuno M. Pedrosa; Ortega Martorell, Sandra; Slotboom, Johannes; Knecht, Urspeter; Arús, Carles; Vellido Alcacena, Alfredo; Julià Sapé, Margarida (I6doc.com, 2016)
    Text en actes de congrés
    Accés obert
    Machine learning has provided, over the last decades, tools for knowledge extraction in complex medical domains. Most of these tools, though, are ad hoc solutions and lack the systematic approach that would be required to ...
  • Instance and feature weighted k-nearest-neighbors algorithm 

    Prat, Gabriel; Belanche Muñoz, Luis Antonio (I6doc.com, 2016)
    Text en actes de congrés
    Accés obert
    We present a novel method that aims at providing a more stable selection of feature subsets when variations in the training process occur. This is accomplished by using an instance-weighting process -assigning different ...
  • Physics and machine learning: Emerging paradigms 

    Martín Guerrero, José; Lisboa, Paulo J G; Vellido Alcacena, Alfredo (I6doc.com, 2016)
    Text en actes de congrés
    Accés obert
    Current research in Machine Learning (ML) combines the study of variations on well-established methods with cutting-edge breakthroughs based on completely new approaches. Among the latter, emerging paradigms from Physics ...

Mostra'n més