Electronic Support (ES) operations concern themselves with the ability to search for, intercept, track and classify threat emitters. Modern radar systems in turn aim to operate undetected by intercept receivers. These radar systems maintain Low Probability of Intercept (LPI) by utilizing low power emissions, coded waveforms, wideband operation, narrow beamwidths and evasive scan patterns without compromising accuracy and resolution. The term LPI refers to the small chance or likelihood of intercept actually occurring. The complexity and degrees of freedom available to modern radar place a high demand on ES systems to provide detailed and accurate real-time information. Intercept alone is not sufficient and this study focusses on the detection, feature extraction (parameter estimation) and classification (using Fuzzy ARTMAP), of the Pilot Mk3 LPI radar. Fuzzy ARTMAP is a cognitive neural method combining fuzzy logic and Adaptive Resonance Theory (ART) to create categories of class prototypes to be classified. Fuzzy ARTMAP systems are formed by self-organizing neural architectures that are able to rapidly learn and classify both discreet and continuous input patterns. To evaluate the suitability of a given ES intercept receiver against a particular LPI radar, the LPI performance factor is defined by combining the radar range, intercept receiver range and sensitivity equations. The radar wants to force an opposing intercept receiver into its range envelope. On the contrary, the intercept receiver would ideally want to operate outside the specified radar detection range to avoid being detected by the radar. The Maximum Likelihood (ML) detector developed for this study is capable of detecting the Pilot Mk3 radar, as it allows sufficient integration gain for detection beyond the radar maximum range. The accuracy of parameter estimation in an intercept receiver is of great importance, as it has a direct impact on the accuracy of the classification stage. Among the various potentially useful radar parameters, antenna rotation rate, transmit frequency, frequency sweep and sweep repetition frequency were used to classify the Pilot Mk3 radar. Estimation of these parameters resulted in very clear clustering of parameter data that distinguish the Pilot Mk3 radar. The estimated radar signal parameters are well separated to the point that there is no overlap of features. If the detector is able to detect an intercepted signal it will be able to make accurate estimates of these parameters. The Fuzzy ARTMAP classifier is capable of classifying the radar modes of the Pilot Mk3 LPI radar. Correct Classification Decisions (CCD) of 100% are easily achieved for a variety of classifier configurations. Classifier training is quite efficient as good generalisation between input and output spaces is achieved from a training dataset comprising only 5% of the total dataset. If any radar is LPI, there must be a consideration for the radar as well as the opposing intercept receiver. Calculating the LPI performance factor is a useful tool for such an evaluation. The claim that a particular radar is LPI against any intercept receiver is too broad to be insightful. This also holds for an intercept receiver claiming to have 100% Probability of Intercept (POI) against any radar. AFRIKAANS : Elektroniese ondersteuningsoperasies het ten doel om uitsendings van bedreigings te soek, te onderskep, te volg en ook te klassifiseer. Moderne radarstelsels probeer op hulle beurt om hul eie werk te verrig sonder om onderskep te word. Hierdie tipe radarstelsels handhaaf ’n Lae Waarskynlikheid van Onderskepping (LWO) d.m.v. lae senderdrywing, geënkodeerde golfvorms, wyebandfrekwensiegebruik, noue antennabundels en vermydende antennasoekpatrone. Hierdie eienskappe veroorsaak dat ’n LWO radar nie akkuraatheid en resolusie prysgee nie. Die term LWO verwys na die skrale kans of waarskynlikheid van onderskepping deur ’n ontvanger wat die radar se gedrag probeer naspeur. Die komplekse seinomgewing en vele grade van vryheid beskikbaar vir ’n LWO-radar, stel baie hoë eise aan onderskeppingsontvangers om gedetaileerde en akkurate inligting in reële tyd te lewer. Die ondersoek van LWO-radaronderskepping op sy eie is nie voldoende nie. Hierdie studie beskou die deteksie, parameter-estimasie asook klassifikasie (m.b.v. Fuzzy ARTMAP) van die Pilot Mk3 LWO-radar as ’n probleem in die geheel. Fuzzy ARTMAP is ’n kognitiewe neurale metode wat fuzzy-logika en Aanspasbare Resonante Teorie (ART) kombineer om kategorieë of klassifikasieprototipes te vorm en hulle te klassifiseer. Fuzzy ARTMAP stelsels bestaan uit selfvormende neurale komponente wat diskrete asook kontinue insette vinnig kan leer en klassifiseer. Om die geskiktheid van enige onderskeppingsontvanger te bepaal word ’n LWO-werkverrigtingsyfer gedefinieer. Hierdie werkverrigtingsyfer kombineer beide radar- en onderskeppings ontvanger vergelykings vir operasionele reikafstand en sensitiwiteit. Die radar beoog om die onderskeppingsontvanger tot binne sy eie reikafstand in te forseer om die ontvangerplatform op te spoor. Die onderskeppingsontvanger wil daarenteen op ’n veilige afstand (verder as die radarbereik) bly, en nogsteeds die radar se uitsendings onderskep. ’n Maksimale Waarskynlikheid (MW) detektor is ontwikkel wat die Pilot Mk3- radargolfvorms kan opspoor, met voldoende integrasie-aanwins vir betroubare deteksie en wat veel verder strek as die radarreikafstand. Akkurate radarparameterestimasie is ’n baie belangrike funksie in ’n onderskeppingsontvanger aangesien dit ’n direkte implikasie het vir die akkuraatheid van die klassifikasiefunksie. Vanuit ’n wye verskeidenheid van relevante radar parameters word estimasies van antennadraaitempo, senderfrekwensie, frekwensieveegbandwydte en veegherhalingstempo gebruik om die Pilot Mk3-radar te klassifiseer. Die estimasie van hierdie parameters is duidelik gegroepeer met geen oorvleuling om moontlike verwarring te voorkom. Indien die detektor deteksies verklaar, volg die estimasiefunksie met baie akkurate waardes van radarparameters. Die Fuzzy ARTMAP-klassifiseerder wat ontwikkel is vir hierdie studie beskik oor die vermoë om die Pilot Mk3 LWO-radar te klassifiseer. Korrekte Klassifikasiebesluite (KKB) van 100% is moontlik vir ’n verskeidenheid klassifiseerderverstellings. Die klassifiseerder behaal ’n goeie veralgemening van in- en uitset ruimtes, en die leer- (of oefen-) roetines is baie effektief met so min as 5% van die volle datastel. Enige radarstelsel wat roem op LWO moet sowel die radar as ’n moontlike onderskeppingsontvanger in gelyke maat beskou. Die LWO- werkverrigtingsyfer verskaf ’n handige maatstaf vir sulke evaluasies. Om bloot te eis dat ’n radar LWO-eienskappe teenoor enige onderskeppingsontvanger het, is te algemeen en nie insiggewend nie. Dieselfde geld vir ’n onderskeppingsontvanger wat 100% (of totale) onderskepping kan verrig teenoor enige radar. Copyright / Dissertation (MEng)--University of Pretoria, 2012. / Electrical, Electronic and Computer Engineering / unrestricted

No description available.

Abstract One of the major problems in computational biology is the inability of existing classification models to incorporate expanding and new domain knowledge. This problem of static classification models is addressed in this thesis by the introduction of incremental learning for problems in bioinformatics. The tools which have been developed are applied to the problem of classifying proteins into a number of primary and putative families. The importance of this type of classification is of particular relevance due to its role in drug discovery programs and the benefit it lends to this process in terms of cost and time saving. As a secondary problem, multi–class classification is also addressed. The standard approach to protein family classification is based on the creation of committees of binary classifiers. This one-vs-all approach is not ideal, and the classification systems presented here consists of classifiers that are able to do all-vs-all classification. Two incremental learning techniques are presented. The first is a novel algorithm based on the fuzzy ARTMAP classifier and an evolutionary strategy. The second technique applies the incremental learning algorithm Learn++. The two systems are tested using three datasets: data from the Structural Classification of Proteins (SCOP) database, G-Protein Coupled Receptors (GPCR) database and Enzymes from the Protein Data Bank. The results show that both techniques are comparable with each other, giving classification abilities which are comparable to that of the single batch trained classifiers, with the added ability of incremental learning. Both the techniques are shown to be useful to the problem of protein family classification, but these techniques are applicable to problems outside this area, with applications in proteomics including the predictions of functions, secondary and tertiary structures, and applications in genomics such as promoter and splice site predictions and classification of gene microarrays.

Classification of terrain cover from satellite radar imagery represents an area of considerable current interest and research. Most satellite sensors used for land applications are of the imaging type. They record data in a variety of spectral channels and at a variety of ground resolutions. Spectral pattern recognition refers to classification procedures utilizing pixel-by-pixel spectral information as the basis for automated land cover classification. A number of methods have been developed in the past to classify pixels [resolution cells] from multispectral imagery to a priori given land cover categories. Their ability to provide land cover information with high classification accuracies is significant for work where accurate and reliable thematic information is needed. The current trend towards the use of more spectral bands on satellite instruments, such as visible and infrared imaging spectrometers, and finer pixel and grey level resolutions will offer more precise possibilities for accurate identification. But as the complexity of the data grows, so too does the need for more powerful tools to analyse them. It is the major objective of this study to analyse the capabilities and applicability of the neural pattern recognition system, called fuzzy ARTMAP, to generate high quality classifications of urban land cover using remotely sensed images. Fuzzy ARTMAP synthesizes fuzzy logic and Adaptive Resonance Theory (ART) by exploiting the formal similarity between the computations of fuzzy subsethood and the dynamics of category choice, search and learning. The paper describes design features, system dynamics and simulation algorithms of this learning system, which is trained and tested for classification (8 a priori given classes) of a multispectral image of a Landsat-5 Thematic Mapper scene (270 x 360 pixels) from the City of Vienna on a pixel-by-pixel basis. Fuzzy ARTMAP performance is compared with that of an error-based learning system based upon the multi-layer perceptron, and the Gaussian maximum likelihood classifier as conventional statistical benchmark on the same database. Both neural classifiers outperform the conventional classifier in terms of classification accuracy. Fuzzy ARTMAP leads to out-of-sample classification accuracies, very close to maximum performance, while the multi-layer perceptron - like the conventional classifier - shows difficulties to distinguish between some land use categories. (authors' abstract) / Series: Discussion Papers of the Institute for Economic Geography and GIScience

En los últimos años las lenguas electrónicas se han convertido en una excelente alternativa a los métodos tradicionales de análisis para el control de los procesos y productos, entre otros, en el ámbito agroalimentario. Se trata de sistemas que, mediante técnicas electroquímicas, como la potenciometría o la voltametría combinadas con herramientas de análisis multivariante, son capaces de clasificar muestras y cuantificar sus parámetros fisicoquímicos. Su funcionamiento se basa en la utilización de sensores de sensibilidad cruzada, lo que permite medir muestras en las que existan interferencias entre los distintos compuestos que la integran. En la actualidad la mayoría de los métodos empleados para la determinación de las propiedades fisicoquímicas son destructivos. El diseño de sistemas de medida no destructivos es un reto. Pero además de preservar la integridad de las muestras analizadas, las nuevas técnicas analíticas deben tener un bajo coste y un funcionamiento sencillo, no dependiente de mano de obra cualificada. Para el análisis de los datos se suele utilizar técnicas de reconocimiento de patrones no supervisadas, como es el Análisis de Componentes Principales (PCA). Pero en muchas ocasiones, es conveniente realizar análisis supervi-sado, donde, las categorías de las muestras están predefinidas y la finalidad es comprobar si es posible conseguir un sistema que sea capaz de clasificar adecuadamente muestras nuevas que entra en el sistema de medida. Uno de los métodos más utilizados para realizar una clasificación de la muestras con técnicas supervisadas son las redes neuronales artificiales (RNA). Existen diversos tipos de redes neuronales, una de las más conocidas y utilizadas es la denominada Perceptrón multicapa. El entrenamiento de esta red consiste en fijar los pesos de cada una de las neuronas. Este tipo de red neuronal, ha comprobado su utilidad en múltiples aplicaciones con lenguas electrónicas, pero también ha demostrado sus limitaciones, que vienen d / Garrigues Baixauli, J. (2013). Implementación de algoritmos de redes neuronales artificiales de tipo Fuzzy Artmap y Multilayer Feed Fordward con dispositivos electrónicos programables en sistemas de lenguas electrónicas para la clasificación de muestras y determinación de parámetros fisicoquímicos [Tesis doctoral]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/30219

El objetivo general del trabajo de esta tesis doctoral, como su nombre indica, consiste en desarrollar, probar y evaluar diferentes estrategias orientadas a superar las limitaciones que actualmente hacen que los sistemas de olfato electrónico (SDOE) no se utilicen en aplicaciones reales, fundamentalmente en aplicaciones relacionadas con la industria agroalimentaria, las cuales requiere urgentemente un equipo de estás características en su planta de producción. Dicho objetivo puede ser subdividido en tres grandes ejes, cuyos resultados han permitido solventar algunas limitaciones importantes de los SDOE:(1) Diseño, construcción y evaluación de un prototipo para la detección de crecimiento fúngico en productos de bollería industrial. En este primer estudio se ha podido determinar el grado de efectividad de diferentes algoritmos de selección de variables que han permitido incrementar espectacularmente la fiabilidad en la respuesta de un sistema de olfato electrónico dedicado a la detección fúngica en bollería industrial. La metodología desarrollada para este objetivo es de fácil implementación en cualquier otra aplicación de los sistemas de olfato electrónico.(2) Desarrollo de un sistema de detección de trazas de benceno en un flujo de CO2.Este segundo trabajo se ha orientado a estudiar las posibilidades reales de incremento en la sensibilidad de los sensores semiconductores comerciales en una aplicación real, con interferentes y bajo una atmósfera nada propicia al funcionamiento de este tipo de dispositivos como es el dióxido de carbono.Para ello se han evaluado diseños alternativos basados en un proceso de concentración y desorción térmica, escogiendo el que mejor resultado podía ofrecer y evaluando así hasta que punto puede ser incrementada la sensibilidad (y selectividad) de un sistema de olfato electrónico mediante esta filosofía.Con el diseño propuesto se ha conseguido incrementar la sensibilidad del equipo en un factor de 100 a 500 veces, siendo además inmune a interferentes en concentraciones órdenes de magnitud por encima del volátil a detectar, el benceno.(3) Diseño y construcción de un sistema de modulación de flujo con vistas a incrementar la selectividad de los sensores semiconductores comerciales frente a diferentes especies gaseosas. Este tercer objetivo incluye tanto la construcción como la evaluación del prototipo que ha permitido comprobar que esta estrategia puede ser aplicada de forma genérica a cualquier sistema de olfato electrónico permitiendo el incremento de selectividad de cualquier sensor semiconductor de óxido de estaño.La combinación de las tres estrategias anteriores, que han sido probadas por separado con gran éxito, debería permitir minimizar las limitaciones actuales de los SDOE.⁻ Organización de la memoria:El capitulo dos trata sobre la detección de hongos en productos de bollería industrial con un Sistema de Olfato Electrónico. Esta aplicación sirve de excusa para comprobar como el acoplar un sistema de selección de variables puede mejorar ostensiblemente el funcionamiento de este tipo de instrumentos en aplicaciones agroalimentarias.El capítulo tres describe los trabajos realizados con un prototipo desarrollado para una aplicación industrial real. En él se ensayan diferentes técnicas de pre-concentración, se determinan tanto la configuración como los modos de operación óptimos, y se evalúa la eficacia del prototipo en una aplicación real. La tercera estrategia ideada para incrementar la efectividad de los sistemas de olfato electrónico es descrita enteramente en el capítulo 4. En él se detallan los entresijos del diseño y acoplamiento de un sistema de modulación de flujo a un SDOE y los resultados que se han obtenido. En el capítulo 5 se detallan las conclusiones obtenidas tras la realización de los estudios descritos en los capítulos 2, 3 y 4, apuntando hacia donde deben continuar los esfuerzos en las líneas de investigación tratadas en esta tesis doctoral. / The general goal of the work of this doctoral thesis, as its name indicates, consists of developing, proving and to evaluate different strategies to improve the limitations that at the moment do that the Electronic Nose System are not used in real applications, fundamentally in applications related to the agro-alimentary industry, which requires an equipment urgently in the production plant. This objective can be subdivided in three great axes, whose results have allowed to resolve some important limitations of the Electronic Nose: (1) Design, construction and evaluation of a prototype for the fungal detection in the industrial bakery products. In this first study it has been possible to determine the degree of effectiveness of different algorithms of variables selection that have allowed spectacularly to increase the reliability in the answer of a Electronic Nose System dedicated to the fungal detection in industrial baker's. The methodology developed for this objective is of easy implementation in any other application of the Electronic Nose system.(2) Development of a system of detection of benzene in a CO2 flow. This second work has been oriented to study the real possibilities of increase in the sensitivity of commercial the semiconducting sensors in a real application, with interferentes and under an atmosphere not at all it causes to the operation of this type of devices as it is carbon dioxide. For it alternative designs based on a concentration process and thermal desorption have been evaluated, choosing the one that better result could offer and thus evaluating until point can be increased the sensitivity (and selectivity) of a Electronic Nose System by means of this philosophy. With the proposed design one has been able to increase the sensitivity of the equipment in a factor of 100 to 500 times, being in addition immune to interferentes in concentrations orders of magnitude over the volatile one to detect, the benzene. (3) Design and construction of a flow modulation system with views to increase the selectivity of commercial the semiconducting sensors front to different gaseous species. This third objective includes so much the construction as the evaluation of the prototype that has allowed to verify that this strategy can be applied of generic form to any system of electronic sense of smell allowing the increase of selectivity of any tin semiconducting oxide sensor. The combination of the three previous strategies, that have been proven separately with great success, would have to allow to diminish the present limitations of a Electronic Nose.Organization:Two chapters deals on the detection of fungi in products of industrial baker's shop with a electronic nose system. This application serves as excuse to verify as connecting a system of variables selection that can improve the operation of this type of instruments in agro-alimentary applications. Chapter three describes the works made with a prototype developed for real an industrial application. Different techniques from pre-concentration are tried, they determine so much the configuration as the optimal ways of operation, and the effectiveness of the prototype in a real application is evaluated. The third devised strategy to increase the effectiveness of the systems of electronic sense of smell is described entirely in chapter 4. In him to the mysteries of the design and connection of a system of modulation of flow to a SDOE are detailed and the results that have been obtained. In chapter 5 the conclusions obtained after the accomplishment of the studies described in chapters 2 are detailed, 3 and 4, aiming towards where they must continue the efforts in the lines of investigation treated in this doctoral thesis.

Submitted by JACQUELINE SANTOS CHAVES null (jac_sc@yahoo.com) on 2016-08-19T20:04:09Z No. of bitstreams: 1 Jacqueline S. Chaves.pdf: 1795331 bytes, checksum: 9c7a177018aa3a98f7cb4a90da94c904 (MD5) / Approved for entry into archive by Juliano Benedito Ferreira (julianoferreira@reitoria.unesp.br) on 2016-08-23T19:46:42Z (GMT) No. of bitstreams: 1 chaves_js_me_ilha.pdf: 1795331 bytes, checksum: 9c7a177018aa3a98f7cb4a90da94c904 (MD5) / Made available in DSpace on 2016-08-23T19:46:42Z (GMT). No. of bitstreams: 1 chaves_js_me_ilha.pdf: 1795331 bytes, checksum: 9c7a177018aa3a98f7cb4a90da94c904 (MD5) Previous issue date: 2016-07-29 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / As técnicas para o monitoramento de falhas em estruturas têm se tornado cada vez mais importantes principalmente por seus benefícios quanto à maior segurança de vida e por auxiliarem as empresas responsáveis em construir edifícios, pontes e estruturas em geral a diminuírem seus custos com a manutenção das mesmas. Deste modo, a fim de desenvolver uma forma eficiente para a identificação e caracterização de falhas estruturais, esta dissertação tem por objetivo demonstrar uma aplicação de Redes Neurais Artificiais (RNAs) como uma técnica de monitoramento da integridade estrutural (SHM) para tal problema. Através de um modelo matemático de equações diferenciais ordinárias para a representação de uma estrutura predial, será desenvolvida uma RNA ARTMAP Fuzzy por ser uma rede flexível e estável em relação à sua habilidade em se adaptar às mudanças imprevistas do ambiente externo, para identificar tais falhas. / The techniques for failures monitoring in mechanical engineering structures have become increasingly important especially for its benefits as the largest life-security and assist the responsible companies for build buildings, bridges and structures in general to lower their costs to maintenance of them. Thus, in order to develop an efficient way for the identification and characterization of structural failures, this work aims to demonstrate an application of Artificial Neural Networks (ANN) as a monitoring technique of structural health monitoring (SHM) for this problem. Through a dynamic model for the representation of a building structure, Fuzzy ARTMAP ANN will be developed to be a flexible and stable network with respect to its ability to adapt to unexpected changes in the external environment to identify such failures.

Orientador: Fábio Roberto Chavarette / Resumo: As técnicas para o monitoramento de falhas em estruturas têm se tornado cada vez mais importantes principalmente por seus benefícios quanto à maior segurança de vida e por auxiliarem as empresas responsáveis em construir edifícios, pontes e estruturas em geral a diminuírem seus custos com a manutenção das mesmas. Deste modo, a fim de desenvolver uma forma eficiente para a identificação e caracterização de falhas estruturais, esta dissertação tem por objetivo demonstrar uma aplicação de Redes Neurais Artificiais (RNAs) como uma técnica de monitoramento da integridade estrutural (SHM) para tal problema. Através de um modelo matemático de equações diferenciais ordinárias para a representação de uma estrutura predial, será desenvolvida uma RNA ARTMAP Fuzzy por ser uma rede flexível e estável em relação à sua habilidade em se adaptar às mudanças imprevistas do ambiente externo, para identificar tais falhas. / Abstract: The techniques for failures monitoring in mechanical engineering structures have become increasingly important especially for its benefits as the largest life-security and assist the responsible companies for build buildings, bridges and structures in general to lower their costs to maintenance of them. Thus, in order to develop an efficient way for the identification and characterization of structural failures, this work aims to demonstrate an application of Artificial Neural Networks (ANN) as a monitoring technique of structural health monitoring (SHM) for this problem. Through a dynamic model for the representation of a building structure, Fuzzy ARTMAP ANN will be developed to be a flexible and stable network with respect to its ability to adapt to unexpected changes in the external environment to identify such failures. / Mestre

The Fuzzy–ARTMAP (FAM) algorithm has been proven to be one of the premier neural network architectures for classification problems. FAM can learn on line and is usually faster than other neural network approaches. Nevertheless the learning time of FAM can slow down considerably when the size of the training set increases into the hundreds of thousands. In this dissertation we apply data partitioning and network partitioning to the FAM algorithm in a sequential and parallel setting to achieve better convergence time and to efficiently train with large databases (hundreds of thousands of patterns). We implement our parallelization on a Beowulf clusters of workstations. This choice of platform requires that the process of parallelization be coarse grained. Extensive testing of all the approaches is done on three large datasets (half a million data points). One of them is the Forest Covertype database from Blackard and the other two are artificially generated Gaussian data with different percentages of overlap between classes. Speedups in the data partitioning approach reached the order of the hundreds without having to invest in parallel computation. Speedups on the network partitioning approach are close to linear on a cluster of workstations. Both methods allowed us to reduce the computation time of training the neural network in large databases from days to minutes. We prove formally that the workload balance of our network partitioning approaches will never be worse than an acceptable bound, and also demonstrate the correctness of these parallelization variants of FAM.

This research focuses on the learning of a class of behaviors defined as high-level behaviors. High-level behaviors are defined here as behaviors that can be executed using a sequence of identifiable behaviors. Represented by low-level contexts, these behaviors are known a priori to learning and can be modeled separately by a knowledge engineer. The learning task, which is achieved by observing an expert within simulation, then becomes the identification and representation of the low-level context sequence executed by the expert. To learn this sequence, this research proposes FAMTILE - the Fuzzy ARTMAP / Template-Based Interpretation Learning Engine. This algorithm attempts to achieve this learning task by constructing rules that govern the low-level context transitions made by the expert. By combining these rules with models for these low-level context behaviors, it is hypothesized that an intelligent model for the expert can be created that can adequately model his behavior. To evaluate FAMTILE, four testing scenarios were developed that attempt to achieve three distinct evaluation goals: assessing the learning capabilities of Fuzzy ARTMAP, evaluating the ability of FAMTILE to correctly predict expert actions and context choices given an observation, and creating a model of the expert's behavior that can perform the high-level task at a comparable level of proficiency.