Empirical evidence for ultrametric structure in multi-layer perceptron error surfaces

Combinatorial optimization problems share an interesting property with spin glass systems in that their state spaces can exhibit ultrametric structure. We use sampling methods to analyse the error surfaces of feedforward multi-layer perceptron neural networks learning encoder problems. The third order statistics of these points of attraction are examined and found to be arranged in a highly ultrametric way. This is a unique result for a finite, continuous parameter space. The implications of this result are discussed.

Learning the dynamics of embedded clauses

Recent work by Siegelmann has shown that the computational power of recurrent neural networks matches that of Turing Machines. One important implication is that complex language classes (infinite languages with embedded clauses) can be represented in neural networks. Proofs are based on a fractal encoding of states to simulate the memory and operations of stacks. In the present work, it is shown that similar stack-like dynamics can be learned in recurrent neural networks from simple sequence prediction tasks. Two main types of network solutions are found and described qualitatively as dynamical systems: damped oscillation and entangled spiraling around fixed points. The potential and limitations of each solution type are established in terms of generalization on two different context-free languages. Both solution types constitute novel stack implementations - generally in line with Siegelmann's theoretical work - which supply insights into how embedded structures of languages can be handled in analog hardware.

Virtual simulation of the postsurgical cosmetic outcome in patientswith pectus excavatum

Pectus excavatum is the most common congenital deformity of the anterior chest wall, in which several ribs and the sternum grow abnormally. Nowadays, the surgical correction is carried out in children and adults through Nuss technic. This technic has been shown to be safe with major drivers as cosmesis and the prevention of psychological problems and social stress. Nowadays, no application is known to predict the cosmetic outcome of the pectus excavatum surgical correction. Such tool could be used to help the surgeon and the patient in the moment of deciding the need for surgery correction. This work is a first step to predict postsurgical outcome in pectus excavatum surgery correction. Facing this goal, it was firstly determined a point cloud of the skin surface along the thoracic wall using Computed Tomography (before surgical correction) and the Polhemus FastSCAN (after the surgical correction). Then, a surface mesh was reconstructed from the two point clouds using a Radial Basis Function algorithm for further affine registration between the meshes. After registration, one studied the surgical correction influence area (SCIA) of the thoracic wall. This SCIA was used to train, test and validate artificial neural networks in order to predict the surgical outcome of pectus excavatum correction and to determine the degree of convergence of SCIA in different patients. Often, ANN did not converge to a satisfactory solution (each patient had its own deformity characteristics), thus invalidating the creation of a mathematical model capable of estimating, with satisfactory results, the postsurgical outcome

Comparative analysis between statistical and artificial intelligence models in business failure prediction

A growing number of predicting corporate failure models has emerged since 60s. Economic and social consequences of business failure can be dramatic, thus it is not surprise that the issue has been of growing interest in academic research as well as in business context. The main purpose of this study is to compare the predictive ability of five developed models based on three statistical techniques (Discriminant Analysis, Logit and Probit) and two models based on Artificial Intelligence (Neural Networks and Rough Sets). The five models were employed to a dataset of 420 non-bankrupt firms and 125 bankrupt firms belonging to the textile and clothing industry, over the period 2003–09. Results show that all the models performed well, with an overall correct classification level higher than 90%, and a type II error always less than 2%. The type I error increases as we move away from the year prior to failure. Our models contribute to the discussion of corporate financial distress causes. Moreover it can be used to assist decisions of creditors, investors and auditors. Additionally, this research can be of great contribution to devisers of national economic policies that aim to reduce industrial unemployment.

Comparative Analysis between Statistical and Artificial Intelligence Models in Business Failure Prediction

A growing number of predicting corporate failure models has emerged since 60s. Economic and social consequences of business failure can be dramatic, thus it is not surprise that the issue has been of growing interest in academic research as well as in business context. The main purpose of this study is to compare the predictive ability of five developed models based on three statistical techniques (Discriminant Analysis, Logit and Probit) and two models based on Artificial Intelligence (Neural Networks and Rough Sets). The five models were employed to a dataset of 420 non-bankrupt firms and 125 bankrupt firms belonging to the textile and clothing industry, over the period 2003–09. Results show that all the models performed well, with an overall correct classification level higher than 90%, and a type II error always less than 2%. The type I error increases as we move away from the year prior to failure. Our models contribute to the discussion of corporate financial distress causes. Moreover it can be used to assist decisions of creditors, investors and auditors. Additionally, this research can be of great contribution to devisers of national economic policies that aim to reduce industrial unemployment.

Automatic assignment of absolute configuration from 1D NMR data

Opposite enantiomers exhibit different NMR properties in the presence of an external common chiral element, and a chiral molecule exhibits different NMR properties in the presence of external enantiomeric chiral elements. Automatic prediction of such differences, and comparison with experimental values, leads to the assignment of the absolute configuration. Here two cases are reported, one using a dataset of 80 chiral secondary alcohols esterified with (R)-MTPA and the corresponding 1H NMR chemical shifts and the other with 94 13C NMR chemical shifts of chiral secondary alcohols in two enantiomeric chiral solvents. For the first application, counterpropagation neural networks were trained to predict the sign of the difference between chemical shifts of opposite stereoisomers. The neural networks were trained to process the chirality code of the alcohol as the input, and to give the NMR property as the output. In the second application, similar neural networks were employed, but the property to predict was the difference of chemical shifts in the two enantiomeric solvents. For independent test sets of 20 objects, 100% correct predictions were obtained in both applications concerning the sign of the chemical shifts differences. Additionally, with the second dataset, the difference of chemical shifts in the two enantiomeric solvents was quantitatively predicted, yielding r2 0.936 for the test set between the predicted and experimental values.

Sistema de apoio à decisão para a classificação de risco em projectos de tecnologias de informação

O trabalho que a seguir se apresenta tem como objectivo descrever a criação de um modelo que sirva de suporte a um sistema de apoio à decisão sobre o risco inerente à execução de projectos na área das Tecnologias de Informação (TI) recorrendo a técnicas de mineração de dados. Durante o ciclo de vida de um projecto, existem inúmeros factores que contribuem para o seu sucesso ou insucesso. A responsabilidade de monitorizar, antever e mitigar esses factores recai sobre o Gestor de Projecto. A gestão de projectos é uma tarefa difícil e dispendiosa, consome muitos recursos, depende de numerosas variáveis e, muitas vezes, até da própria experiência do Gestor de Projecto. Ao ser confrontado com as previsões de duração e de esforço para a execução de uma determinada tarefa, o Gestor de Projecto, exceptuando a sua percepção e intuição pessoal, não tem um modo objectivo de medir a plausibilidade dos valores que lhe são apresentados pelo eventual executor da tarefa. As referidas previsões são fundamentais para a organização, pois sobre elas são tomadas as decisões de planeamento global estratégico corporativo, de execução, de adiamento, de cancelamento, de adjudicação, de renegociação de âmbito, de adjudicação externa, entre outros. Esta propensão para o desvio, quando detectada numa fase inicial, pode ajudar a gerir melhor o risco associado à Gestão de Projectos. O sucesso de cada projecto terminado foi qualificado tendo em conta a ponderação de três factores: o desvio ao orçamentado, o desvio ao planeado e o desvio ao especificado. Analisando os projectos decorridos, e correlacionando alguns dos seus atributos com o seu grau de sucesso o modelo classifica, qualitativamente, um novo projecto quanto ao seu risco. Neste contexto o risco representa o grau de afastamento do projecto ao sucesso. Recorrendo a algoritmos de mineração de dados, tais como, árvores de classificação e redes neuronais, descreve-se o desenvolvimento de um modelo que suporta um sistema de apoio à decisão baseado na classificação de novos projectos. Os modelos são o resultado de um extensivo conjunto de testes de validação onde se procuram e refinam os indicadores que melhor caracterizam os atributos de um projecto e que mais influenciam o risco. Como suporte tecnológico para o desenvolvimento e teste foi utilizada a ferramenta Weka 3. Uma boa utilização do modelo proposto possibilitará a criação de planos de contingência mais detalhados e uma gestão mais próxima para projectos que apresentem uma maior propensão para o risco. Assim, o resultado final pretende constituir mais uma ferramenta à disposição do Gestor de Projecto.

Controlo tolerante a falhas de um canal de rega

A presente dissertação pretende conceber e implementar um sistema de controlo tolerante a falhas, no canal experimental de rega da Universidade de Évora, utilizando um modelo implementado em MATLAB/SIMULINK®. Como forma de responder a este desafio, analisaram-se várias técnicas de diagnóstico de falhas, tendo-se optado por técnicas baseadas em redes neuronais para o desenvolvimento de um sistema de detecção e isolamento de falhas no canal de rega, sem ter em conta o tipo de sistema de controlo utilizado. As redes neuronais foram, assim, os processadores não lineares utilizados e mais aconselhados em situações onde exista uma abundância de dados do processo, porque aprendem por exemplos e são suportadas por teorias estatísticas e de optimização, focando não somente o processamento de sinais, como também expandindo os horizontes desse processamento. A ênfase dos modelos das redes neuronais está na sua dinâmica, na sua estabilidade e no seu comportamento. Portanto, o trabalho de investigação do qual resultou esta Dissertação teve como principais objectivos o desenvolvimento de modelos de redes neuronais que representassem da melhor forma a dinâmica do canal de rega, de modo a obter um sistema de detecção de falhas que faça uma comparação entre os valores obtidos nos modelos e no processo. Com esta diferença de valores, da qual resultará um resíduo, é possível desenvolver tanto o sistema de detecção como de isolamento de falhas baseados nas redes neuronais, possibilitando assim o desenvolvimento dum sistema de controlo tolerante a falhas, que engloba os módulos de detecção, de isolamento/diagnóstico e de reconfiguração do canal de rega. Em síntese, na Dissertação realizada desenvolveu-se um sistema que permite reconfigurar o processo em caso de ocorrência de falhas, melhorando significativamente o desempenho do canal de rega.

Short-term electricity prices forecasting in a competitive market by a hybrid PSO-ANFIS approach

In this paper, a novel hybrid approach is proposed for electricity prices forecasting in a competitive market, considering a time horizon of 1 week. The proposed approach is based on the combination of particle swarm optimization and adaptive-network based fuzzy inference system. Results from a case study based on the electricity market of mainland Spain are presented. A thorough comparison is carried out, taking into account the results of previous publications, to demonstrate its effectiveness regarding forecasting accuracy and computation time. Finally, conclusions are duly drawn. (C) 2012 Elsevier Ltd. All rights reserved.

Application of adaptive neuro-fuzzy inference for wind power short-term forecasting

The increased integration of wind power into the electric grid, as nowadays occurs in Portugal, poses new challenges due to its intermittency and volatility. Hence, good forecasting tools play a key role in tackling these challenges. In this paper, an adaptive neuro-fuzzy inference approach is proposed for short-term wind power forecasting. Results from a real-world case study are presented. A thorough comparison is carried out, taking into account the results obtained with other approaches. Numerical results are presented and conclusions are duly drawn. (C) 2011 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

ANN-Based LMP forecasting in a distribution network with large penetration of DG

In recent years, power systems have experienced many changes in their paradigm. The introduction of new players in the management of distributed generation leads to the decentralization of control and decision-making, so that each player is able to play in the market environment. In the new context, it will be very relevant that aggregator players allow midsize, small and micro players to act in a competitive environment. In order to achieve their objectives, virtual power players and single players are required to optimize their energy resource management process. To achieve this, it is essential to have financial resources capable of providing access to appropriate decision support tools. As small players have difficulties in having access to such tools, it is necessary that these players can benefit from alternative methodologies to support their decisions. This paper presents a methodology, based on Artificial Neural Networks (ANN), and intended to support smaller players. In this case the present methodology uses a training set that is created using energy resource scheduling solutions obtained using a mixed-integer linear programming (MIP) approach as the reference optimization methodology. The trained network is used to obtain locational marginal prices in a distribution network. The main goal of the paper is to verify the accuracy of the ANN based approach. Moreover, the use of a single ANN is compared with the use of two or more ANN to forecast the locational marginal price.

QSAR modeling of antitubercular activity of diverse organic compounds

Tuberculosis (TB) is a worldwide infectious disease that has shown over time extremely high mortality levels. The urgent need to develop new antitubercular drugs is due to the increasing rate of appearance of multi-drug resistant strains to the commonly used drugs, and the longer durations of therapy and recovery, particularly in immuno-compromised patients. The major goal of the present study is the exploration of data from different families of compounds through the use of a variety of machine learning techniques so that robust QSAR-based models can be developed to further guide in the quest for new potent anti-TB compounds. Eight QSAR models were built using various types of descriptors (from ADRIANA.Code and Dragon software) with two publicly available structurally diverse data sets, including recent data deposited in PubChem. QSAR methodologies used Random Forests and Associative Neural Networks. Predictions for the external evaluation sets obtained accuracies in the range of 0.76-0.88 (for active/inactive classifications) and Q(2)=0.66-0.89 for regressions. Models developed in this study can be used to estimate the anti-TB activity of drug candidates at early stages of drug development (C) 2011 Elsevier B.V. All rights reserved.

Decision support system with incomplete and domain incoherent information management

Many of the most common human functions such as temporal and non-monotonic reasoning have not yet been fully mapped in developed systems, even though some theoretical breakthroughs have already been accomplished. This is mainly due to the inherent computational complexity of the theoretical approaches. In the particular area of fault diagnosis in power systems however, some systems which tried to solve the problem, have been deployed using methodologies such as production rule based expert systems, neural networks, recognition of chronicles, fuzzy expert systems, etc. SPARSE (from the Portuguese acronym, which means expert system for incident analysis and restoration support) was one of the developed systems and, in the sequence of its development, came the need to cope with incomplete and/or incorrect information as well as the traditional problems for power systems fault diagnosis based on SCADA (supervisory control and data acquisition) information retrieval, namely real-time operation, huge amounts of information, etc. This paper presents an architecture for a decision support system, which can solve the presented problems, using a symbiosis of the event calculus and the default reasoning rule based system paradigms, insuring soft real-time operation with incomplete, incorrect or domain incoherent information handling ability. A prototype implementation of this system is already at work in the control centre of the Portuguese Transmission Network.

LMP based bid formation for virtual power players operating in smart grids

Power system organization has gone through huge changes in the recent years. Significant increase in distributed generation (DG) and operation in the scope of liberalized markets are two relevant driving forces for these changes. More recently, the smart grid (SG) concept gained increased importance, and is being seen as a paradigm able to support power system requirements for the future. This paper proposes a computational architecture to support day-ahead Virtual Power Player (VPP) bid formation in the smart grid context. This architecture includes a forecasting module, a resource optimization and Locational Marginal Price (LMP) computation module, and a bid formation module. Due to the involved problems characteristics, the implementation of this architecture requires the use of Artificial Intelligence (AI) techniques. Artificial Neural Networks (ANN) are used for resource and load forecasting and Evolutionary Particle Swarm Optimization (EPSO) is used for energy resource scheduling. The paper presents a case study that considers a 33 bus distribution network that includes 67 distributed generators, 32 loads and 9 storage units.

ANN based support for distributed energy resources scheduling in smart grids

The future scenarios for operation of smart grids are likely to include a large diversity of players, of different types and sizes. With control and decision making being decentralized over the network, intelligence should also be decentralized so that every player is able to play in the market environment. In the new context, aggregator players, enabling medium, small, and even micro size players to act in a competitive environment, will be very relevant. Virtual Power Players (VPP) and single players must optimize their energy resource management in order to accomplish their goals. This is relatively easy to larger players, with financial means to have access to adequate decision support tools, to support decision making concerning their optimal resource schedule. However, the smaller players have difficulties in accessing this kind of tools. So, it is required that these smaller players can be offered alternative methods to support their decisions. This paper presents a methodology, based on Artificial Neural Networks (ANN), intended to support smaller players’ resource scheduling. The used methodology uses a training set that is built using the energy resource scheduling solutions obtained with a reference optimization methodology, a mixed-integer non-linear programming (MINLP) in this case. The trained network is able to achieve good schedule results requiring modest computational means.