Marcelino Martínez

BELM: Bayesian Extreme Learning Machine

The theory of extreme learning machine (ELM) has become very popular on the last few years. ELM is a new approach for learning the parameters of the hidden layers of a multilayer neural network (as the multilayer perceptron or the radial basis function neural network). Its main advantage is the lower computational cost, which is especially relevant when dealing with many patterns defined in a high-dimensional space. This brief proposes a Bayesian approach to ELM, which presents some advantages over other approaches: it allows the introduction of a priori knowledge; obtains the confidence intervals (CIs) without the need of applying methods that are computationally intensive, e.g., bootstrap; and presents high generalization capabilities. Bayesian ELM is benchmarked against classical ELM in several artificial and real datasets that are widely used for the evaluation of machine learning algorithms. Achieved results show that the proposed approach produces a competitive accuracy with some additional advantages, namely, automatic production of CIs, reduction of probability of model overfitting, and use of a priori knowledge.

A novel approach to introducing adaptive filters based on the LMS algorithm and its variants

This paper presents a new approach to introducing adaptive filters based on the least-mean-square (LMS) algorithm and its variants in an undergraduate course on digital signal processing. Unlike other filters currently taught to undergraduate students, these filters are nonlinear and time variant. This proposal introduces adaptive filtering in the context of a linear time-invariant system using a real problem. In this way, introducing adaptive filters using concepts already familiar to the students motivates their interest through practical application. The key point for this simplification is that the input to the filter is constant so that the adaptive filter becomes linear. Therefore, a complete arsenal of mathematical tools, already known by the students, is available to analyze the performance of the filters and obtain the key parameters to adaptive filters, e.g., speed of convergence and stability. Several variants of the basic LMS algorithm are described the same way.

Application of ARMA modeling to the improvement of weight estimations in fruit sorting and grading machinery

Accurate weighting of pieces in different sorts of conveyor belts or articulated chains at fast speed is a key feature in many industrial processes. This paper presents a procedure to improve the performance, whether increasing speed or accuracy, of the load-cell-based weighting subsystem in a fruit sorting and grading machine to achieve an accuracy of /spl plusmn/1 gram at a speed of 20 fruits per seconds. The proposed solution includes a signal preprocessing based on a previous ARMA modeling of the weighting subsystem response plus a power line-noise removal and a simple sample averaging in the plateau. The procedure has been tested off-line using real signals acquired from a prototype machine.

Survival prediction in patients undergoing ischemic cardiopathy

The ischemic cardiopathy is the main cause of death in developed countries. New improved drugs and therapies have appeared last years. However, the interventionist strategy and the most powerful drugs may have complications, and hence, it is very important to know the risk of death associated with patients during their stay in the hospital, or in the next six months. Thus, it is possible to tune the best treatment for each individual patient. In this framework, the use of artificial neural networks is proposed with a double objective: survival prediction and the extraction of the parameters with best predictive capabilities. A cohort of 691 patients treated in the Hospital Clínic, in Barcelona (Spain) during the period 2006–08 was used for this study. The obtained results show the good prediction capabilities of neural models when compared with classical models (logistic regression) and decision trees. Moreover, neural models reduced the number of relevant variables for the prediction from 134 to only 36.

Adaptive algorithms robust to impulsive noise with low computational cost using Order Statistic

Abstract In this paper a family of adaptive algorithms robust to impulsive noise and with low computational cost are presented. Unlike other approaches, no cost functions or filtering of the gradient are considered in order to update the filter coefficients. Its initial basis is the basic LMS algorithm and its sign-error variant. The proposed algorithms can be considered as some sign-error variants of the LMS algorithm. The algorithms are successfully tested in terms of accuracy and convergence in a standard system identification simulation in which an impulsive noise is present. Simulations show that they improve the performance of LMS variants that are robust to impulsive noise.

A Family of Adaptive Algorithms Robust to Impulsive Noise

This paper presents a new approach to the development of a family of adaptive algorithms that are robust to impulsive noise. Unlike other approaches, no cost functions or filtering of the gradient are considered in order to update the filter coefficients. Basically, the algorithm takes into account the distance between the absolute errors and the median of the absolute values of the most recent errors committed by the adaptive filter. The proposed family of algorithms can be considered as a sign-error variant of the LMS algorithm. The proposed adaptive algorithm is successfully tested in terms of accuracy and convergence in a system identification simulation in which an impulsive noise is present.