Neural networks complemented with genetic algorithms and fuzzy systems for predicting nitrogenous effluent variables in wastewater treatment plants

This work focuses on the prediction of the two main nitrogenous variables that describe the water quality at the effluent of a Wastewater Treatment Plant. We have developed two kind of Neural Networks architectures based on considering only one output or, in the other hand, the usual five effluent variables that define the water quality: suspended solids, biochemical organic matter, chemical organic matter, total nitrogen and total Kjedhal nitrogen. Two learning techniques based on a classical adaptative gradient and a Kalman filter have been implemented. In order to try to improve generalization and performance we have selected variables by means genetic algorithms and fuzzy systems. The training, testing and validation sets show that the final networks are able to learn enough well the simulated available data specially for the total nitrogen

(No)competition in the Spanish retailing gasoline market: a variance filter approach

Various methodologies in economic literature have been used to analyse the international hydrocarbon retail sector. Nevertheless at a Spanish level these studies are much more recent and most conclude that generally there is no effective competition present in this market, regardless of the approach used. In this paper, in order to analyse the price levels in the Spanish petrol market, our starting hypothesis is that in uncompetitive markets the prices are higher and the standard deviation is lower. We use weekly retail petrol price data from the ten biggest Spanish cities, and apply Markov chains to fill the missing values for petrol 95 and diesel, and we also employ a variance filter. We conclude that this market demonstrates reduced price dispersion, regardless of brand or city.

BAW notch filter for WiMAX applications

Als darrers anys la necessitat de connectar-se a internet des de qualsevol lloc s’ha incrementat exponencialment sobretot de manera inalàmbrica. Degut al finit espectre radioelèctric es tendeix a aprofitar totes les franges freqüencials d’aquest convivint diferents sistemes en franges properes podent induir-se interferències mútuament. Per evitar aquestes interferències es requereix de filtres a tots els dispositius els quals aïllin un sistema del adjacent. En aquest projecte es dóna una solució al cas concret de la convivència entre els sistemes Wi-Fi y WiMAX eliminant la banda Wi-Fi interferent en sistemes WiMAX. Aquesta solució consisteix en el disseny d’un filtre banda eliminada d’ordre 3 implementat mitjançant tecnologia BAW a partir de l’estructura y especificacions d’un filtre comercial. A més també es fa un petit estudi per veure si seria interessant una millora en els processos de fabricació del filtre per part del fabricant.

Implementación y entrenamiento de un modelo clasificatorio de red neural sobre la base de datos IGBADAT para la clasificación de las rocas basálticas de acuerdo a las clases del sistema de clasificación tradicional de Yoder and Tiller.

Aquest treball vol implementar un projecte de mineria de dades en l'àrea de la petrologia ígnia, especialitat englobada dins la geologia clàssica.

Efficient learning of reactive robot behaviors with a Neural-Q_learning approach

The purpose of this paper is to propose a Neural-Q_learning approach designed for online learning of simple and reactive robot behaviors. In this approach, the Q_function is generalized by a multi-layer neural network allowing the use of continuous states and actions. The algorithm uses a database of the most recent learning samples to accelerate and guarantee the convergence. Each Neural-Q_learning function represents an independent, reactive and adaptive behavior which maps sensorial states to robot control actions. A group of these behaviors constitutes a reactive control scheme designed to fulfill simple missions. The paper centers on the description of the Neural-Q_learning based behaviors showing their performance with an underwater robot in a target following task. Real experiments demonstrate the convergence and stability of the learning system, pointing out its suitability for online robot learning. Advantages and limitations are discussed

Semi-online neural-Q_leaming for real-time robot learning

Reinforcement learning (RL) is a very suitable technique for robot learning, as it can learn in unknown environments and in real-time computation. The main difficulties in adapting classic RL algorithms to robotic systems are the generalization problem and the correct observation of the Markovian state. This paper attempts to solve the generalization problem by proposing the semi-online neural-Q_learning algorithm (SONQL). The algorithm uses the classic Q_learning technique with two modifications. First, a neural network (NN) approximates the Q_function allowing the use of continuous states and actions. Second, a database of the most representative learning samples accelerates and stabilizes the convergence. The term semi-online is referred to the fact that the algorithm uses the current but also past learning samples. However, the algorithm is able to learn in real-time while the robot is interacting with the environment. The paper shows simulated results with the "mountain-car" benchmark and, also, real results with an underwater robot in a target following behavior

Volatility smile extrapolation with an artificial neural network

I use a multi-layer feedforward perceptron, with backpropagation learning implemented via stochastic gradient descent, to extrapolate the volatility smile of Euribor derivatives over low-strikes by training the network on parametric prices.

Doppler ultrasound in the measurement of pulse wave velocity: agreement with the Complior method

Aortic stiffness is an independent predictor factor for cardiovascular risk. Different methods for determining pulse wave velocity (PWV) are used, among which the most common are mechanical methods such as SphygmoCor or Complior, which require specific devices and are limited by technical difficulty in obtaining measurements. Doppler guided by 2D ultrasound is a good alternative to these methods. We studied 40 patients (29 male, aged 21 to 82 years) comparing the Complior method with Doppler. Agreement of both devices was high (R = 0.91, 0.84-0.95, 95% CI). The reproducibility analysis revealed no intra-nor interobserver differences. Based on these results, we conclude that Doppler ultrasound is a reliable and reproducible alternative to other established methods for themeasurement of aortic PWV

Cannabinoid addiction: behavioral models and neural correlates

The use of cannabis sativa preparations as recreational drugs can be traced back to the earliest civilizations. However, animal models of cannabinoid addiction allowing the exploration of neural correlates of cannabinoid abuse have been developed only recently. We review these models and the role of the CB1 cannabinoid receptor, the main target of natural cannabinoids, and its interaction with opioid and dopamine transmission in reward circuits. Extensive reviews on the molecular basis of cannabinoid action are available elsewhere (Piomelli et al., 2000;Schlicker and Kathmann, 2001).

What can spike train distances tell us about the neural code?

Time scale parametric spike train distances like the Victor and the van Rossum distancesare often applied to study the neural code based on neural stimuli discrimination.Different neural coding hypotheses, such as rate or coincidence coding,can be assessed by combining a time scale parametric spike train distance with aclassifier in order to obtain the optimal discrimination performance. The time scalefor which the responses to different stimuli are distinguished best is assumed to bethe discriminative precision of the neural code. The relevance of temporal codingis evaluated by comparing the optimal discrimination performance with the oneachieved when assuming a rate code.We here characterize the measures quantifying the discrimination performance,the discriminative precision, and the relevance of temporal coding. Furthermore,we evaluate the information these quantities provide about the neural code. Weshow that the discriminative precision is too unspecific to be interpreted in termsof the time scales relevant for encoding. Accordingly, the time scale parametricnature of the distances is mainly an advantage because it allows maximizing thediscrimination performance across a whole set of measures with different sensitivitiesdetermined by the time scale parameter, but not due to the possibility toexamine the temporal properties of the neural code.

The HP-filter in cross-country comparisons

Many empirical studies of business cycles have followed the practise ofapplying the Hodrick-Prescott filter for cross-country comparisons. Thestandard procedure is to set the weight \lambda, which determines the'smoothness' of the trend equal to 1600. We show that if this value isused for against common wisdom about business cycles. As an example, weshow that the long recession occurred inSpain between 1975 and 1985 goesunnotoced by the HP filter. We propose a method for adjusting \lambda byreinterpreting the HP-filter as the solution to a constrained minimizationproblem. We argue that the common practice of fixing \lambda across countriesamounts to chankging the constraints on trend variability across countries.Our proposed method is easy to apply, retains all the virtues of thestandard HP-filter and when applied to Spanish data the results are inthe line with economic historian's view. Applying the method to a numberof OECD countries we find that, with the exception of Spain, Italy andJapan, the standard choice of \lambda=1600 is sensible.

A low dimensional Kalman filter for systems with lagged observables

This note describes how the Kalman filter can be modified to allow for thevector of observables to be a function of lagged variables without increasing the dimensionof the state vector in the filter. This is useful in applications where it is desirable to keepthe dimension of the state vector low. The modified filter and accompanying code (whichnests the standard filter) can be used to compute (i) the steady state Kalman filter (ii) thelog likelihood of a parameterized state space model conditional on a history of observables(iii) a smoothed estimate of latent state variables and (iv) a draw from the distribution oflatent states conditional on a history of observables.

Validation procedures in radiological diagnostic models. Neural network and logistic regression

The objective of this paper is to compare the performance of twopredictive radiological models, logistic regression (LR) and neural network (NN), with five different resampling methods. One hundred and sixty-seven patients with proven calvarial lesions as the only known disease were enrolled. Clinical and CT data were used for LR and NN models. Both models were developed with cross validation, leave-one-out and three different bootstrap algorithms. The final results of each model were compared with error rate and the area under receiver operating characteristic curves (Az). The neural network obtained statistically higher Az than LR with cross validation. The remaining resampling validation methods did not reveal statistically significant differences between LR and NN rules. The neural network classifier performs better than the one based on logistic regression. This advantage is well detected by three-fold cross-validation, but remains unnoticed when leave-one-out or bootstrap algorithms are used.

Regulation of GABA(A) and glutamate receptor expression, synaptic facilitation and long-term potentiation in the hippocampus of prion mutant mice

Background: Prionopathies are characterized by spongiform brain degeneration, myoclonia, dementia, and periodic electroencephalographic (EEG) disturbances. The hallmark of prioniopathies is the presence of an abnormal conformational isoform (PrP(sc)) of the natural cellular prion protein (PrP(c)) encoded by the Prnp gene. Although several roles have been attributed to PrP(c), its putative functions in neuronal excitability are unknown. Although early studies of the behavior of Prnp knockout mice described minor changes, later studies report altered behavior. To date, most functional PrP(c) studies on synaptic plasticity have been performed in vitro. To our knowledge, only one electrophysiological study has been performed in vivo in anesthetized mice, by Curtis and coworkers. They reported no significant differences in paired-pulse facilitation or LTP in the CA1 region after Schaffer collateral/commissural pathway stimulation. Principal Findings: Here we explore the role of PrP(c) expression in neurotransmission and neural excitability using wild-type, Prnp -/- and PrP(c)-overexpressing mice (Tg20 strain). By correlating histopathology with electrophysiology in living behaving mice, we demonstrate that both Prnp -/- mice but, more relevantly Tg20 mice show increased susceptibility to KA, leading to significant cell death in the hippocampus. This finding correlates with enhanced synaptic facilitation in paired-pulse experiments and hippocampal LTP in living behaving mutant mice. Gene expression profiling using Illumina microarrays and Ingenuity pathways analysis showed that 129 genes involved in canonical pathways such as Ubiquitination or Neurotransmission were co-regulated in Prnp -/- and Tg20 mice. Lastly, RT-qPCR of neurotransmission-related genes indicated that subunits of GABA(A) and AMPA-kainate receptors are co-regulated in both Prnp -/- and Tg20 mice. Conclusions/Significance: Present results demonstrate that PrP(c) is necessary for the proper homeostatic functioning of hippocampal circuits, because of its relationships with GABA(A) and AMPA-Kainate neurotransmission. New PrP(c) functions have recently been described, which point to PrP(c) as a target for putative therapies in Alzheimer's disease. However, our results indicate that a "gain of function" strategy in Alzheimer's disease, or a "loss of function" in prionopathies, may impair PrP(c) function, with devastating effects. In conclusion, we believe that present data should be taken into account in the development of future therapies.

Podocalyxin is a novel polysialylated neural adhesion protein with multiple roles in neural development and synapse formation

Neural development and plasticity are regulated by neural adhesion proteins, including the polysialylated form of NCAM (PSA-NCAM). Podocalyxin (PC) is a renal PSA-containing protein that has been reported to function as an anti-adhesin in kidney podocytes. Here we show that PC is widely expressed in neurons during neural development. Neural PC interacts with the ERM protein family, and with NHERF1/2 and RhoA/G. Experiments in vitro and phenotypic analyses of podxl-deficient mice indicate that PC is involved in neurite growth, branching and axonal fasciculation, and that PC loss-of-function reduces the number of synapses in the CNS and in the neuromuscular system. We also show that whereas some of the brain PC functions require PSA, others depend on PC per se. Our results show that PC, the second highly sialylated neural adhesion protein, plays multiple roles in neural development.