Journal articles on the topic 'Gradiente descendentes'

Was ist neu? Stand der Dinge Sowohl im klinischen Bereich als auch in der Forschung etabliert sich zunehmend das Konzept, die Aorta als Gesamtorgan zu betrachten, das sich von der Aortenklappe bis zur Bifurkation erstreckt. Aortenklappenstenose Aktuell laufen erste randomisierte Studien zur Ausweitung der Indikation zum interventionellen Aortenklappenersatz (TAVI, TAVR) hin zu einem Patientenkollektiv mit niedrigem OP-Risiko. Des Weiteren hat die interventionelle Versorgung der „Low flow, low gradient“-Aortenklappenstenose eine Aufwertung im Empfehlungsgrad erhalten und die interventionelle Valve-in-valve-Prozedur Eingang in den klinischen Alltag bei der individualisierten Patientenversorgung gefunden. Aortenklappeninsuffizienz Kathetergestützt implantierte Klappenprothesen, die dediziert zur Behandlung der isolierten Aortenklappeninsuffizienz entwickelt werden, befinden sich für dieses Krankheitsbild in der klinischen Prüfung. Erkrankungen der Aorta ascendens und descendens Während das Screening von Aortenaneurysmen zur ambulanten Routine wird, sind Studien zum risikostratifizierten Follow-up bei zunehmend endovaskulär versorgten Patienten notwendig, um eine bedarfsgerechte Nachsorge zu ermöglichen.

The designing and modeling of delta wing is one of the interesting topics. A number of researchers has contributed different works on modeling the same. This paper comes out with a new delta wing modeling with the inclusion of optimization concept. The obtained data from the investigation is integrated and given as the input to the classifier for predicting the drag and lift coefficients. This paper uses neural network (NN) classifier for predicting the drag and lift coefficients. Moreover, the weight of the NN is optimized using a proposed genetic algorithm. After the implementation, the performance of proposed model is compared to other conventional methods like individual adaptive genetic algorithm (IAGA-NN), deterministic adaptive genetic algorithm (DAGA-NN), self-adaptive genetic algorithm (SAGA-NN), genetic algorithm (GA-NN), gradient descendent (GD-NN), and Levenberg masquerade (LM-NN), respectively, in terms of drag and lift coefficient.

Resumo: Este estudo, sob o olhar da Sociofonética, apresenta uma análise acústica das variantes dos sons de r-forte na fala de descendentes italianos da cidade de Rio do Sul, Alto Vale do Itajaí - SC. Nessa região, encontramos a variante tepe ([ɾ]), como uma das possibilidades de produção nesse contexto. Participaram da pesquisa seis representantes da comunidade, divididos em três faixas etárias: 20-50, 51-70 e acima de 70, do sexo feminino e masculino, que narraram suas histórias de vida em entrevistas de fala semiespontânea. O corpus foi formado por 147 itens lexicais que apresentavam as variantes de róticos em onset silábico inicial ou medial de palavra. Com base nos resultados acústicos, foi verificada a gradiência dos dados e foi constatado que os descendentes italianos da comunidade estudada produzem, como r-forte, diferentes variantes, além do tepe ([ɾ]). Fatores, como idade, identidade e região de origem, parecem influenciar na variedade de róticos produzidos. Foi observada uma diminuição na produção do tepe na dimensão diageracional, o que aponta para um estudo em tempo aparente.Palavras-chave: róticos; Alto Vale do Itajaí (SC); variação; Sociofonética.Abstract: This study, under the perspective of Sociofonética, presents an acoustic analysis of the variants of the sounds of r-forte in the speech of Italian descendants of the city of Rio do Sul, Alto Vale do Itajaí - SC. In this region, we find the variant tepe ([ɾ]), as one of the production possibilities in this context. Six community representatives participated in the research, divided into three age groups: 20-50, 51-70 and over 70, female and male, who narrated their life stories in semi-spontaneous speech interviews. The corpus was formed by 147 lexical items that presented the rhotic variants in initial or medial syllabic onset of the word. Based on the acoustic results, the gradient of the data was verified and the hypothesis was confirmed that the Italian descendants of the studied community produce, as r-forte, different variants, in addition to the tepe ([ɾ]). Factors, such as age, identity and region of origin, seem to influence the variety of rotics produced. A decrease in the production of tepe in the diagerational dimension was observed, which points to a study in apparent time.Keywords: rothics (R-sounds); Alto Vale do Itajaí (SC); Variation; Sociophonetics.

Debranning was applied to durum wheat to the study the relationship between kernel shape and size, and ash and mineral distribution having implications for semolina yield. To this aim four durum wheat genotypes carried out over three environments were selected to determine the morphological and yield traits as well as the distribution along the kernel of the ash, macro- (Na, K, P, Ca, and Mg), and micro-elements (Mn, Fe, Cu, Zn, and Mo). A descendent ash gradient within the kernel reflects the decreases in the minerals that occurred during debranning. Perciasacchi with high seed weight (TKW) and greater thickness followed by Cappelli showed a more uniform distribution of ash content along the kernels. High r Pearson coefficient (p < 0.01) showed an inverse relationship between thickness and ash decay. Since thickness was strongly affected by the genotype, it could represent a useful trait for breeding programs to predict the milling quality.

Este artículo describe el diseño, implementación y prueba de la herramienta denominada UV-SRNA-PDA (simulador de redes neuronales artificiales de la Universidad del Valle para una agenda digital programable), que está orientada a la simulación de redes neuronales artificiales y a la identificación de procesos industriales complejos. Esta aplicación trabaja sobre una agenda digital programable (PDA) Palm T5 usando un sistema de adquisición de datos diseñado para tal fin. Dos tipos de redes neuronales fueron implementados: el perceptrón y el perceptrón multicapa (MLP) usando como algoritmos de aprendizaje los siguientes: propagación hacia atrás, gradiente descendente, velocidad de aprendizaje variable y momentum. Las pruebas se realizaron sobre plantas de primer y segundo orden (sólo esta última es reportada en este artículo), obteniendo su modelo neuronal y validando sus resultados en dos plataformas conocidas: MATLAB y UV-SRNA 2.0 (versión de la UV-SRNA-PDA para PC). Estas pruebas arrojaron un error de entrenamiento promedio de 5.62 × 10-3 ± 3.55 × 10-4 y un error de validación promedio de 4.56 × 10-3 ± 5.95 × 10-4. En ambos casos, los resultados son mejores o comparables con los de las otras herramientas de simulación. Sin embargo, el tiempo de entrenamiento típico en UV-SRNA-PDA fue de 900 s en comparación con los 3 s para MATLAB y 8 s para la UV-SRNA 2.0

Un tema de interés sobre los vehículos aéreos no tripulados es la determinación de trayectorias apropiadas que le permitan moverse de una posición inicial a una posición objetivo, garantizando que la trayectoria sea segura, es decir, que no existan riesgos de colisión. En este artículo se presentan dos técnicas de planificación de trayectorias y una estrategia de evasión de obstáculos para cuadricópteros, ambas técnicas funcionales sobre ambientes tridimensionales con obstáculos estáticos o dinámicos restringidos a velocidades constantes. Las técnicas funcionan sobre un ambiente modelado con planos que generan un campo de potencial artificial. La primera técnica está basada en puntos móviles que interconectan la posición inicial con la meta, posteriormente cada punto se mueve hacia zonas libres de la influencia de los obstáculos siguiendo el campo potencial, lo que hace que se encuentren caminos libres de obstáculos. La segunda técnica usa el concepto de zonas seguras, el cual se utiliza como criterio para actualizar la posición de los puntos. Adicionalmente, se propone una metodología de evasión de obstáculos dinámicos, que consiste en transformar el problema de planeación de trayectorias con obstáculos dinámicos al caso de planeación de trayectorias con obstáculos estáticos mediante un proceso de análisis de las posibilidades de colisión. Los resultados muestran que estas técnicas superan los inconvenientes de los algoritmos basados en gradiente descendente como mínimos locales y oscilaciones inestables.

Neste trabalho, propomos um esquema de controle adaptativo de fluxos de tráfego de redes de computadores baseado no modelo fuzzy Takagi-Sugeno (TSK). O esquema de controle proposto evita a ocorrência de congestionamento utilizando o modelo fuzzy TSK para prever adaptativamente o tamanho da fila no buffer nos enlaces. Com o objetivo de ajustar dinamicamente os parâmetros do modelo fuzzy TSK, desenvolvemos um algoritmo de treinamento adaptativo para o mesmo. Os parâmetros das partes de premissa e das partes conseqüentes do modelo são atualizados na primeira etapa do algoritmo de treinamento desenvolvido. Já a segunda etapa consiste de um algoritmo de refinamento dos parâmetros do modelo baseado em gradiente descendente. A eficiência do preditor proposto é avaliada com o uso de traços de tráfego reais em comparação a outros preditores. A partir dos parâmetros do modelo fuzzy TSK, deduzimos uma expressão para a taxa da fonte de tráfego de modo a minimizar a variância do tamanho de fila no buffer e que é parte essencial do nosso esquema de controle. O controle de congestionamento proposto é então aplicado em um cenário de rede composto de vários nós alimentados com fluxos de tráfego real. Comparações realizadas com outros métodos de controle de congestionamento demonstram que o controle de congestionamento proposto provê menores taxas de perdas, assim como a manutenção da taxa de ocupação no buffer abaixo do valor desejado.

El objetivo de la presente investigación fue actualizar el mapa de climas del estado de Jalisco de acuerdo con la clasificación climática Köppen-García, con base en imágenes raster climáticas de alta resolución (90 m) y un ajuste por correspondencia climas-vegetación potencial. Se trabajó con datos climáticos diarios, mensuales y anuales de 154 estaciones meteorológicas del estado de Jalisco y 151 estaciones de estados vecinos. Los datos corresponden a la serie 1961-2014 en el caso de las estaciones de Jalisco y la serie 1961-2010 para estados vecinos. Con esta información se obtuvieron imágenes raster para la clasificación del clima con el método Köppen-García, mediante procesos de interpolación implementados con el método Anusplin. En el sistema Idrisi y utilizando rutinas de álgebra de mapas, se obtuvo el mapa de climas de Jalisco a nivel de tipo y subtipo climático. Dicho mapa se corrigió atendiendo a la correspondencia de los climas obtenidos con la vegetación potencial según COTECOCA. Los resultados mostraron correspondencia de 84% entre climas y tipos de vegetación y sólo 16% de los polígonos climáticos fueron corregidos. Se obtuvieron 29 variantes climáticas, de las cuales, los climas secos o semiáridos representan 13.78% del territorio estatal, cálido húmedos Aw el 21.77%, semicálidos húmedos A(C)w el 4.45%, semicálidos húmedos (A)C(w) el 41.67%, templados húmedos Cw el 18.33% y fríos E, el 0.001%. En orden descendente, el número de climas por región fue: Región Sur 21, Costa Sierra Occidental 19, Sierra de Amula 18, Norte 14, Sureste 13, Costa Sur 12, Altos Sur 11, Valles 10, Lagunas 9, Centro 8, Altos Norte 7 y Ciénega 7. Se concluye que el estado de Jalisco posee un gradiente térmico de cálido a frío, incluyendo la semicálida y templada y por un gradiente de humedad la semiárida hasta la subhúmeda monzónica. Los autores consideran que los mapas climáticos obtenidos, pueden ser útiles en aplicaciones para diversas áreas, como la agricultura, edafología, ecología y ciencias ambientales.

Ureteral balloon catheters and ureteral stents are implanted in large quantities on a daily basis. They are the suspected cause for about a quarter of all the nosocomial infections, which lead to approx. 20,000 deaths in Germany alone. To fight these infections, catheters should be made antibacterial. A technique for an antibacterial coating of catheters exhibiting an aspect ratio of up to 200 consists of a thin silver layer, which is deposited out of an aqueous solution, which is followed by a second step: chemical vapor deposition (CVD) of an organic polymeric film, which moderates the release rate of silver ions. The main concern of the second step is the longitudinal evenness of the film. For tubes with one opening as balloon catheters, this issue can be solved by applying a descendent temperature gradient from the opening to the end of the catheter. An alternative procedure can be applied to commercially available ureteral stents, which exhibit small drainage openings in their middle. The same CVD as before leads to a longitudinal homogeneity of about ±10%—at very low costs. This deposition can be modeled using viscous flow.

Introducción. Una fístula coronaria se define como la comunicación entre una arteria coronaria y una cámara cardiaca o cualquier segmento de la circulación sistémica o pulmonar. Su incidencia en series angiográficas y población general es muy baja y, en general, cursan de modo asintomático, aunque en raras ocasiones presentan significancia hemodinámica. Su localización en cavidades izquierdas es menos frecuente. Presentación del caso. Paciente masculino de 52 años quien presenta angina de esfuerzo, por lo que fue estratificado de modo invasivo encontrándose una fístula coronaria de la arteria descendente anterior al ventrículo izquierdo. Ante dichos hallazgos, le fue solicitada resonancia magnética nuclear cardiaca como estudio complementario para determinar conducta terapéutica a futuro; sin embargo, el paciente no asistió a controles, ni se le realizó dicho examen. Discusión. Las fístulas coronarias que causan enfermedad arterial coronaria son raras y el drenaje de una fístula coronaria a ventrículo izquierdo es aún más infrecuente. La importancia fisiopatológica de una fístula coronaria está relacionada con el volumen de sangre que fluye y el gradiente de presión a través de la comunicación. La mayoría de fístulas coronarias son diagnosticadas incidentalmente en un cateterismo cardíaco; sin embargo, algunas de estas presentan significancia clínica, siendo sintomáticas y causando complicaciones, por lo que requieren tratamiento a corto plazo. Conclusiones. En algunos casos, las pruebas complementarias iniciales en pacientes con clínica de isquemia miocárdica, realizadas de modo no invasivo, permiten sospechar la presencia de fístulas coronarias. La angiografía coronaria continúa siendo la prueba de mayor precisión diagnostica. Además, se deben tener en cuenta las características anatómicas y fisiológicas para definir si requieren manejo y si este se hará por vía percutánea o quirúrgica.

ABSTRACT A comprehensive, combined, numerical model is presented, based on genetic algorithm (GA) optimization and heat-transfer finite-element computations. The numerical analyses were carried out to evaluate the final cross-linking degree of a medium-voltage electric cable subjected to industrial peroxide reticulation. The difference between numerically predicted and experimentally determined cross-linking degree along the thickness of the insulator is then minimized, when a variable steam-temperature profile along the pipe length is assumed to explain the unexpected undervulcanization of the cable in the internal layers. To minimize the gap between the experimentally determined degree of curing and numerical predictions, a GA optimization is used for best fit, instead of the steepest-descendent, standard least-squares, which is not applicable in this case because the objective function is not analytically known. The cable was supposedly vulcanized under four different conditions. The degree of cross-linking is experimentally obtained by determining the differential scanning calorimetry (DSC) of the nondecomposed peroxide from the external layer to the core of the cable. The proposed GA approach exploits a specifically crafted, zooming strategy, consisting of the subdivision of the population at each iteration into two subgroups, depending on the individual's grade of fitness (elitist strategy). The integrated, numerical, experimental approach allows optimization of the amount of peroxide in the compound and comparison of the performances of different peroxide mixtures. The degree of cross-linking compliance can be obtained as a function of the temperature gradient measured in the steam pipe. Production conditions can be automatically calculated according to the cable parameters, by increasing the quality reliability and reducing the scraps.

Algumas propriedades do biodiesel, como índice de iodo, viscosidade e densidade, podem sofrer variações conforme as estruturas moleculares dos seus ésteres constituintes. O objetivo do pre-sente estudo é avaliar e comparar três métodos de convergência no treinamento supervisionado de redes neurais com arquitetura MLP na predição de propriedades de biodiesel. Os métodos aplicados foram os de BFGS, Gradiente Descendente e Gradiente Conjugado. Dados do LAPQAP e da literatura foram padronizados, organizados e armazenados no Sistema Oleodata, sendo divididos em três partes: 70% para o treinamento da rede, 15% para a fase de validação e 15% para a de teste. As variáveis de entrada foram os percentuais de cada éster de ácido graxo que compõe as amostras de biodiesel, e as variáveis de saída foram o índice de iodo, a viscosidade e a densidade. As seguintes funções de ativação foram previamente testadas: tangente hiperbólica, seno, identidade, exponencial e logística. Além disso, foram fixadas a quantidade de 1000 redes treinadas para cada variação do estudo e 1000 ciclos de treinamento para cada rede. Em seguida, foram variados os métodos de convergência: BFGS, Gradiente Descenden-te e Gradiente Conjugado. Após a obtenção do método de convergência, um novo treinamento foi feito para a otimização da quantidade de neurônios, variando-se de 5 a 15. Ao final de todos os treinamentos a rede com melhores desempenhos foi treinada pelo método de BFGS e apresentava 10 neurônios ocul-tos, função logística em ambas as camadas oculta e de saída, com um coeficiente médio superior a 0,88.Palavras-chave: Parâmetros de Qualidade. Método de BFGS. Método de Gradiente Descendente. Mé-todo de Gradiente Conjugado.EVALUATION OF CONVERGENCE METHODS IN TRAINING OF ARTIFICIAL NEURAL NETWORKS APPLIED TO PREDICTION OF IODINE INDEX, VISCOSITY AND DENSITY IN BIODIESELABSTRACT: Some properties of biodiesel, like iodine index, viscosity and density, can vary with the mo-lecular structures of its esters. The present study evaluates and compares three convergence methods of supervised training of neural networks with MLP architecture on prediction of biodiesel properties. The methods applied were BFGS, Descent Gradient and Conjugate Gradient. Data from LAPQAP and from literature were standardized, organized and stored into the Oleodata System, being divided in three parts: 70% for networks training, 15% for validation phase and 15% for test. The input variables were the per-cent of each fatty ester which compose the biodiesel samples, and the output variables were iodine index, viscosity and density. The following activation functions were previously tested: hyperbolic tangent, sine, identity, exponential and logistic. Moreover, it was fixed the quantity of 1000 networks trained for each va-riation of the study and 1000 cycles of training for each network. Then, convergence methods were varied: BFGS, Descent Gradient and Conjugate Gradient. After selected the best convergence method, a new training was carried out for optimization of neurons quantity, ranging from 5 to 15. At the end of training, the network with best performances was trained by BFGS method and presented 10 hidden neurons, logistic function in both hidden and output layers, with a mean coefficient higher than 0,88.KEYWORDS: Quality Parameters. BFGS Method. Descent Gradient Method. Conjugate Gradient Method.EVALUACIÓN DE MÉTODOS DE CONVERGENCIA EN EL ENTRENAMIENTO DE REDES NEURONALES ARTIFICIALES APLICADAS A LA PREDICCIÓN DEL ÍNDICE DE YODO, LA VISCOSIDAD Y LA DENSIDAD DE BIODIESELRESUMEN: Algunas propiedades del biodiesel, como índice de yodo, viscosidad y densidad, pueden va-riar con las estructuras moleculares de sus ésteres. El presente estudio evalúa y compara tres métodos de convergencia de entrenamiento supervisado de las redes neuronales con la arquitectura MLP en la predicción de las propiedades del biodiesel. Los métodos aplicados fueron BFGS, Gradiente de Descen-so y Gradiente Conjugado. Los datos de LAPQAP y de la literatura fueron estandarizados, organizado y almacenado en el Sistema de Oleodata, siendo divididos en tres partes: 70% para el entrenamiento de redes, 15% para la fase de validación y 15% para el test. Las variables de entrada fueron el porcentaje de cada éster de grasa que componen las muestras de biodiesel, y las variables de salida fueron el índice de yodo, la viscosidad y la densidad. Las siguientes funciones de activación fueron probadas previamente: tangente hiperbólica, seno, identidad, exponencial y logística. Además, se determinó la cantidad de 1.000 redes entrenadas para cada variación del estudio y 1000 ciclos de formación para cada red. Entonces, los métodos de convergencia fueron variados: BFGS, Gradiente de Descenso y Gradiente Conjugado. Después de seleccionar el mejor método de convergencia, un nuevo entrenamiento fue realizado para la optimización de la cantidad de neuronas, que van de 5 a 15. Al final de la formación, la red con las mejores actuaciones fue entrenada por el método BFGS y presentó 10 neuronas ocultas, la función logística en capas encubiertas y de salida, con un coeficiente medio superior a 0,88.PALABRAS CLAVE: Parámetros de Calidad. Método de BFGS. Método del Gradiente de Descenso. Método del Gradiente Conjugado.

AbstractMany laser scanners depend on their mechanical construction to guarantee their measurements accuracy, however, the current computational technologies allow us to improve these measurements by mathematical methods implemented in neural networks. In this article we are going to introduce the current laser scanner technologies, give a description of our 3D laser scanner and adjust their measurement error by a previously trained feed forward back propagation (FFBP) neural network with a Widrow-Hoff weight/bias learning function. A comparative analysis with other learning functions such as the Kohonen algorithm and gradient descendent with momentum algorithm is presented. Finally, computational simulations are conducted to verify the performance and method uncertainty in the proposed system.

A fin de facilitar el diseño curricular así como el proceso de enseñanza-aprendizaje-evaluación,se propone un ordenamiento de las competencias profesionales de la Odontología basado enniveles de complejidad. La secuencia completa según gradiente descendente de complejidad es:perfil profesional, competencia, función, tarea, paso de tarea, movimiento y momento. Ladefinición de competencias y funciones se realiza mediante una matriz de doble entrada, basadaen la relación entre necesidades de salud y funciones requeridas para la atención de dichasnecesidades. El aprendizaje de los distintos niveles de complejidad requiere que el alumnocomprometa todo su potencial cognitivo, afectivo, psicomotor y volitivo. Se examina lanaturaleza cuali-cuantitativa del proceso enseñanza-aprendizaje y se discute la necesidad deemplear tanto una evaluación basada en información objetiva como una ponderaciónintersubjetiva, tal como se expone en el artículo. Se enfatiza la necesidad de que el alumno logreintegrar las relaciones entre los componentes de una competencia y de las competencias entresí, a fin de asegurar que el nuevo graduado posea la capacidad necesaria para solucionar losproblemas que plantea el ejercicio profesional. Finalmente, el artículo presenta ejemplos deenunciados de competencias de acuerdo con los niveles taxonómicos propuestos.

Abstract Background A 47-year old female complaining of exertional dyspnoea (NYHA class III) was admitted at our Cardiology department. She had a history of nodular sclerosis Hodgkin lymphoma (HL), treated with chemo- and radiotherapy, and complicated by post-actinic pneumopathy and cardiopathy. At the age of 39, she had undergone coronary artery bypass grafting with left internal mammal artery (LIMA) to left anterior descendent artery and saphenous vein to obtuse marginal branch, and aortic valve replacement with a mechanical prosthesis due to severe aortic stenosis. Some years later, she had undergone percutaneous stenting of the left main (LM) due to occlusion of the LIMA bypass graft. At admission, the patient was hemodynamically stable, with signs of right-sided congestive heart failure. Both 2D and 3D transthoracic echocardiogram (TTE) showed preserved biventricular function, normal function of the aortic prosthesis, and diffuse calcification of the whole mitral valve apparatus, involving the leaflets, the annulus, the tendinous chords and the anterolateral papillary muscle (Figure Panels A-B), causing severe mitral stenosis (mean gradient 10 mmHg, 3D planimetric area 0.9 cm2, Wilkins score 12) and moderate organic insufficiency (Panel C). The tricuspid valve was also affected, with thickened, hypomobile leaflets, causing mild stenosis (mean gradient 4 mmHg, 3D planimetric area 3.8 cm2) and severe insufficiency (Panel D). Transesophageal echocardiogram (TOE) couldn"t be performed because of actinic oesophagitis. Percutaneous valvuloplasty was contraindicated due to moderate mitral insufficiency, high Wilkins score and a huge amount of calcium affecting the whole valve apparatus but sparing the commissures. The patient was scheduled to PCI on the LM due to intrastent restenosis, but died during the procedure as a consequence of an intrastent massive thrombosis leading to cardiac arrest. Learning points Hodgkin lymphoma survivors are at increased cardiovascular and intraoperative risk. Old radiotherapy protocols for HL may cause severe post-actinic valvular and coronary disease. Post-actinic valvular heart disease often affects aortic and mitral valve more than a decade after irradiation, and may manifest as stenosis, insufficiency or both. Organic regurgitation and stenosis of tricuspid valve are uncommon, but may also occur and lead to worse patient outcome. Despite TOE may bring additional valuable informations in challenging cases, the coexistence of oesophageal sequelae from post-actinic oesophagitis may limit its applicability. TTE is the first line and often the only diagnostic tool available for identifying the characteristic valvular lesions in cancer survivors exposed to radiotherapy. 3D TTE may be particularly useful to identify subtle signs of primary involvement of tricuspid apparatus and quantify the anatomical area of a stenotic tricuspid valve, when severe regurgitation coexists and transvalvular gradients may be unreliable. Abstract P167 Figure

Abstract: The supply chain performance evaluation is a critical activity to continuously improve operations. Literature presents several performance evaluation systems based on multi-criteria methods and artificial intelligence. Among them, the systems based on artificial neural networks (ANN) excel due to their capacity of modeling non-linear relationships between metrics and allowing adaptations to a specific environment by means of historical performance data. These systems’ accuracy depend directly on the adopted training algorithm, and no studies have been found that assess the efficiency of these algorithms when applied to supply chain performance evaluation. In this context, the present study evaluates four ANNs learning methods in order to investigate which one is the most adequate to deal with supply chain evaluation. The algorithms tested were Gradient Descendent Momentum, Levenberg-Marquardt, Quasi-Newton and Scale Conjugate Gradient. The performance metrics were extracted from SCOR®, which is a reference model used worldwide. The random sub-sampling cross-validation method was adopted to find the most adequate topological configuration for each model. A set of 80 topologies was implemented using MATLAB®. The prediction accuracy evaluation was based on the mean square error. For the four level 1 metrics considered, the Levenberg-Marquardt algorithm provided the most precise results. The results of correlation analysis and hypothesis tests reinforce the accuracy of the proposed models. Furthermore, the proposed computational models reached a prediction accuracy higher than previous approaches.

Abstract A 88-year-old female was admitted for an anterior ST-segment elevation myocardial infarction (STEMI). Patient had a history of intermittent chest pain for 2 days with worsening on the day of admission. Electrocardiogram (ECG) at admission showed ST-segment elevation on leads from V2 to V6 and leads DI and aVL. Initial observation on the emergency department was described as unremarkable apart from the chest pain. Aspirin and Ticagrelor loading doses were administered and patient underwent emergent coronary angiography, which showed left anterior descendent artery occlusion after the emergence of second diagonal branch. Coronary angioplasty of this lesion was tried, with a total of 3 drug-eluted stents implantation but with no success as in the end there was no reflow of the artery. Patient was then admitted on cardiac intensive care unit, and on observation at that time there was a remarking holosystolic murmur. Transthoracic echocardiogram showed (apart from left ventricle systolic disfunction with akinesia of the apical segments as well as middle segments of the interventricular septum (IVS) and anterior wall) an apical IVS defect with a left to right shunt with a gradient of around 50mmHg evaluated by Doppler, and no signs of right ventricle overload. Case was promptly discussed with cardiothoracic surgery and it was decided that she was not a candidate to urgent surgical intervention. Patient had an initial evolution in Killip class II, and remained hemodynamically stable for the rest of the admission, having no signs of heart failure on discharge at 17 days later. Serial ETTs during admission and at discharge were similar to the evaluation performed at admission. In the meanwhile, during admission, case was discussed in multidisciplinary heart team with cardiothoracic surgery and interventional cardiology. Given the favourable evolution and comorbidities and frailty of the patient it was decided to adopt a conservative strategy with medical follow-up, only considering intervention if there was worsening of heart failure. Until now, with 4 months follow-up, patient remains in New York Heart Association (NYHA) functional class I. Discussion Post-myocardial infarction ventricular septal defect (VSD) is a complication that, regardless of the treatment strategy has a high mortality rate, especially when patient presents in cardiogenic shock. However, when patient is stable and especially when comorbidities imposes a high interventional risk medical treatment can be an option. So far, this is a successful case of a medically managed post-myocardial infarction VSD. Abstract P862 Figure. Ventricular Septal Defect