Learning real-world stimuli in a neural network with spike-driven synaptic dynamics

We present a model of spike-driven synaptic plasticity inspired by experimental observations and motivated by the desire to build an electronic hardware device that can learn to classify complex stimuli in a semisupervised fashion. During training, patterns of activity are sequentially imposed on the input neurons, and an additional instructor signal drives the output neurons toward the desired activity. The network is made of integrate-and-fire neurons with constant leak and a floor. The synapses are bistable, and they are modified by the arrival of presynaptic spikes. The sign of the change is determined by both the depolarization and the state of a variable that integrates the postsynaptic action potentials. Following the training phase, the instructor signal is removed, and the output neurons are driven purely by the activity of the input neurons weighted by the plastic synapses. In the absence of stimulation, the synapses preserve their internal state indefinitely. Memories are also very robust to the disruptive action of spontaneous activity. A network of 2000 input neurons is shown to be able to classify correctly a large number (thousands) of highly overlapping patterns (300 classes of preprocessed Latex characters, 30 patterns per class, and a subset of the NIST characters data set) and to generalize with performances that are better than or comparable to those of artificial neural networks. Finally we show that the synaptic dynamics is compatible with many of the experimental observations on the induction of long-term modifications (spike-timing-dependent plasticity and its dependence on both the postsynaptic depolarization and the frequency of pre- and postsynaptic neurons).

PSO and neural networks: Optimal combination to solve non linear complex problems

Swarm colonies reproduce social habits. Working together in a group to reach a predefined goal is a social behaviour occurring in nature. Linear optimization problems have been approached by different techniques based on natural models. In particular, Particles Swarm optimization is a meta-heuristic search technique that has proven to be effective when dealing with complex optimization problems. This paper presents and develops a new method based on different penalties strategies to solve complex problems. It focuses on the training process of the neural networks, the constraints and the election of the parameters to ensure successful results and to avoid the most common obstacles when searching optimal solutions.

Topographic mappings and feed-forward neural networks

This thesis is a study of the generation of topographic mappings - dimension reducing transformations of data that preserve some element of geometric structure - with feed-forward neural networks. As an alternative to established methods, a transformational variant of Sammon's method is proposed, where the projection is effected by a radial basis function neural network. This approach is related to the statistical field of multidimensional scaling, and from that the concept of a 'subjective metric' is defined, which permits the exploitation of additional prior knowledge concerning the data in the mapping process. This then enables the generation of more appropriate feature spaces for the purposes of enhanced visualisation or subsequent classification. A comparison with established methods for feature extraction is given for data taken from the 1992 Research Assessment Exercise for higher educational institutions in the United Kingdom. This is a difficult high-dimensional dataset, and illustrates well the benefit of the new topographic technique. A generalisation of the proposed model is considered for implementation of the classical multidimensional scaling (¸mds}) routine. This is related to Oja's principal subspace neural network, whose learning rule is shown to descend the error surface of the proposed ¸mds model. Some of the technical issues concerning the design and training of topographic neural networks are investigated. It is shown that neural network models can be less sensitive to entrapment in the sub-optimal global minima that badly affect the standard Sammon algorithm, and tend to exhibit good generalisation as a result of implicit weight decay in the training process. It is further argued that for ideal structure retention, the network transformation should be perfectly smooth for all inter-data directions in input space. Finally, there is a critique of optimisation techniques for topographic mappings, and a new training algorithm is proposed. A convergence proof is given, and the method is shown to produce lower-error mappings more rapidly than previous algorithms.

Globally optimal learning rates in multilayer neural networks

A method for calculating the globally optimal learning rate in on-line gradient-descent training of multilayer neural networks is presented. The method is based on a variational approach which maximizes the decrease in generalization error over a given time frame. We demonstrate the method by computing optimal learning rates in typical learning scenarios. The method can also be employed when different learning rates are allowed for different parameter vectors as well as to determine the relevance of related training algorithms based on modifications to the basic gradient descent rule.

Exploratory database visualisation: the application and assessment of data and dimensionality reduction

This thesis describes the development of a complete data visualisation system for large tabular databases, such as those commonly found in a business environment. A state-of-the-art 'cyberspace cell' data visualisation technique was investigated and a powerful visualisation system using it was implemented. Although allowing databases to be explored and conclusions drawn, it had several drawbacks, the majority of which were due to the three-dimensional nature of the visualisation. A novel two-dimensional generic visualisation system, known as MADEN, was then developed and implemented, based upon a 2-D matrix of 'density plots'. MADEN allows an entire high-dimensional database to be visualised in one window, while permitting close analysis in 'enlargement' windows. Selections of records can be made and examined, and dependencies between fields can be investigated in detail. MADEN was used as a tool for investigating and assessing many data processing algorithms, firstly data-reducing (clustering) methods, then dimensionality-reducing techniques. These included a new 'directed' form of principal components analysis, several novel applications of artificial neural networks, and discriminant analysis techniques which illustrated how groups within a database can be separated. To illustrate the power of the system, MADEN was used to explore customer databases from two financial institutions, resulting in a number of discoveries which would be of interest to a marketing manager. Finally, the database of results from the 1992 UK Research Assessment Exercise was analysed. Using MADEN allowed both universities and disciplines to be graphically compared, and supplied some startling revelations, including empirical evidence of the 'Oxbridge factor'.

Practical methods of tracking of nonstationary time series applied to real-world data

In this paper, we discuss some practical implications for implementing adaptable network algorithms applied to non-stationary time series problems. Two real world data sets, containing electricity load demands and foreign exchange market prices, are used to test several different methods, ranging from linear models with fixed parameters, to non-linear models which adapt both parameters and model order on-line. Training with the extended Kalman filter, we demonstrate that the dynamic model-order increment procedure of the resource allocating RBF network (RAN) is highly sensitive to the parameters of the novelty criterion. We investigate the use of system noise for increasing the plasticity of the Kalman filter training algorithm, and discuss the consequences for on-line model order selection. The results of our experiments show that there are advantages to be gained in tracking real world non-stationary data through the use of more complex adaptive models.

A novel neural network based system for assessing risks associated with information technology security breaches

Security remains a top priority for organizations as their information systems continue to be plagued by security breaches. This dissertation developed a unique approach to assess the security risks associated with information systems based on dynamic neural network architecture. The risks that are considered encompass the production computing environment and the client machine environment. The risks are established as metrics that define how susceptible each of the computing environments is to security breaches. ^ The merit of the approach developed in this dissertation is based on the design and implementation of Artificial Neural Networks to assess the risks in the computing and client machine environments. The datasets that were utilized in the implementation and validation of the model were obtained from business organizations using a web survey tool hosted by Microsoft. This site was designed as a host site for anonymous surveys that were devised specifically as part of this dissertation. Microsoft customers can login to the website and submit their responses to the questionnaire. ^ This work asserted that security in information systems is not dependent exclusively on technology but rather on the triumvirate people, process and technology. The questionnaire and consequently the developed neural network architecture accounted for all three key factors that impact information systems security. ^ As part of the study, a methodology on how to develop, train and validate such a predictive model was devised and successfully deployed. This methodology prescribed how to determine the optimal topology, activation function, and associated parameters for this security based scenario. The assessment of the effects of security breaches to the information systems has traditionally been post-mortem whereas this dissertation provided a predictive solution where organizations can determine how susceptible their environments are to security breaches in a proactive way. ^

Digital Forensics AI: on Practicality, Optimality, and Interpretability of Digital Evidence Mining Techniques

Digital forensics as a field has progressed alongside technological advancements over the years, just as digital devices have gotten more robust and sophisticated. However, criminals and attackers have devised means for exploiting the vulnerabilities or sophistication of these devices to carry out malicious activities in unprecedented ways. Their belief is that electronic crimes can be committed without identities being revealed or trails being established. Several applications of artificial intelligence (AI) have demonstrated interesting and promising solutions to seemingly intractable societal challenges. This thesis aims to advance the concept of applying AI techniques in digital forensic investigation. Our approach involves experimenting with a complex case scenario in which suspects corresponded by e-mail and deleted, suspiciously, certain communications, presumably to conceal evidence. The purpose is to demonstrate the efficacy of Artificial Neural Networks (ANN) in learning and detecting communication patterns over time, and then predicting the possibility of missing communication(s) along with potential topics of discussion. To do this, we developed a novel approach and included other existing models. The accuracy of our results is evaluated, and their performance on previously unseen data is measured. Second, we proposed conceptualizing the term “Digital Forensics AI” (DFAI) to formalize the application of AI in digital forensics. The objective is to highlight the instruments that facilitate the best evidential outcomes and presentation mechanisms that are adaptable to the probabilistic output of AI models. Finally, we enhanced our notion in support of the application of AI in digital forensics by recommending methodologies and approaches for bridging trust gaps through the development of interpretable models that facilitate the admissibility of digital evidence in legal proceedings.

Directly training spiking neural networks for cyber-physical systems: from supervised to reinforcement learning

Spiking Neural Networks (SNNs) are bio-inspired Artificial Neural Networks (ANNs) utilizing discrete spiking signals, akin to neuron communication in the brain, making them ideal for real-time and energy-efficient Cyber-Physical Systems (CPSs). This thesis explores their potential in Structural Health Monitoring (SHM), leveraging low-cost MEMS accelerometers for early damage detection in motorway bridges. The study focuses on Long Short-Term SNNs (LSNNs), although their complex learning processes pose challenges. Comparing LSNNs with other ANN models and training algorithms for SHM, findings indicate LSNNs' effectiveness in damage identification, comparable to ANNs trained using traditional methods. Additionally, an optimized embedded LSNN implementation demonstrates a 54% reduction in execution time, but with longer pre-processing due to spike-based encoding. Furthermore, SNNs are applied in UAV obstacle avoidance, trained directly using a Reinforcement Learning (RL) algorithm with event-based input from a Dynamic Vision Sensor (DVS). Performance evaluation against Convolutional Neural Networks (CNNs) highlights SNNs' superior energy efficiency, showing a 6x decrease in energy consumption. The study also investigates embedded SNN implementations' latency and throughput in real-world deployments, emphasizing their potential for energy-efficient monitoring systems. This research contributes to advancing SHM and UAV obstacle avoidance through SNNs' efficient information processing and decision-making capabilities within CPS domains.

Aplicação de redes neuronais artificiais à deteção e isolamento de falhas em processos industriais

Trabalho Final de Mestrado para obtenção do grau de Mestre em Engenharia Mecânica

Desenvolvimento do padrão de crescimento urbano: modelação da dinâmica do solo urbano em Andaluzia

Dissertação apresentada como requisito parcial para obtenção do grau de Mestre em Ciência e Sistemas de Informação Geográfica

Classificação e previsão de cardiotocogramas usando algoritmos baseados em redes neuronais

Mestrado em Computação e Instrumentação Médica

Previsão do Diagrama de Carga de Subestações da REN utilizando Redes Neuronais

A presente dissertação apresenta o estudo de previsão do diagrama de carga de subestações da Rede Elétrica Nacional (REN) utilizando redes neuronais, com o intuito de verificar a viabilidade do método utilizado, em estudos futuros. Atualmente, a energia elétrica é um bem essencial e desempenha um papel fundamental, tanto a nível económico do país, como a nível de conforto e satisfação individual. Com o desenvolvimento do setor elétrico e o aumento dos produtores torna-se importante a realização da previsão de diagramas de carga, contribuindo para a eficiência das empresas. Esta dissertação tem como objetivo a utilização do modelo das redes neuronais artificiais (RNA) para criar uma rede capaz de realizar a previsão de diagramas de carga, com a finalidade de oferecer a possibilidade de redução de custos e gastos, e a melhoria de qualidade e fiabilidade. Ao longo do trabalho são utilizados dados da carga (em MW), obtidos da REN, da subestação da Prelada e dados como a temperatura, humidade, vento e luminosidade, entre outros. Os dados foram devidamente tratados com a ajuda do software Excel. Com o software MATLAB são realizados treinos com redes neuronais, através da ferramenta Neural Network Fitting Tool, com o objetivo de obter uma rede que forneça os melhores resultados e posteriormente utiliza-la na previsão de novos dados. No processo de previsão, utilizando dados reais das subestações da Prelada e Ermesinde referentes a Março de 2015, comprova-se que com a utilização de RNA é possível obter dados de previsão credíveis, apesar de não ser uma previsão exata. Deste modo, no que diz respeito à previsão de diagramas de carga, as RNA são um bom método a utilizar, uma vez que fornecem, à parte interessada, uma boa previsão do consumo e comportamento das cargas elétricas. Com a finalização deste estudo os resultados obtidos são no mínimo satisfatórios. Consegue-se alcançar através das RNA resultados próximos aos valores que eram esperados, embora não exatamente iguais devido à existência de uma margem de erro na aprendizagem da rede neuronal.

Entwicklung eines Windleistungprognosemodells zur Verbesserung der Kraftwerkseinsatzplanung

Die Maßnahmen zur Förderung der Windenergie in Deutschland haben wichtige Anstöße zur technologischen Weiterentwicklung geliefert und die Grundlagen für den enormen Anlagenzubau geschaffen. Die installierte Windleistung hat heute eine beachtliche Größenordnung erreicht und ein weiteres Wachstum in ähnlichen Dimensionen ist auch für die nächsten Jahre zu erwarten. Die aus Wind erzeugte elektrische Leistung deckt bereits heute in einigen Netzbereichen die Netzlast zu Schwachlastzeiten. Dies zeigt, dass die Windenergie ein nicht mehr zu vernachlässigender Faktor in der elektrischen Energieversorgung geworden ist. Im Rahmen der Kraftwerkseinsatzplanung sind Betrag und Verlauf der Windleistung des folgenden Tages mittlerweile zu wichtigen und zugleich schwierig zu bestimmenden Variablen geworden. Starke Schwankungen und falsche Prognosen der Windstromeinspeisung verursachen zusätzlichen Bedarf an Regel- und Ausgleichsleistung durch die Systemführung. Das im Rahmen dieser Arbeit entwickelte Prognosemodell liefert die zu erwartenden Windleistungen an 16 repräsentativen Windparks bzw. Gruppen von Windparks für bis zu 48 Stunden im Voraus. Aufgrund von prognostizierten Wetterdaten des deutschen Wetterdienstes (DWD) werden die Leistungen der einzelnen Windparks mit Hilfe von künstlichen neuronalen Netzen (KNN) berechnet. Diese Methode hat gegenüber physikalischen Verfahren den Vorteil, dass der komplexe Zusammenhang zwischen Wettergeschehen und Windparkleistung nicht aufwendig analysiert und detailliert mathematisch beschrieben werden muss, sondern anhand von Daten aus der Vergangenheit von den KNN gelernt wird. Das Prognosemodell besteht aus zwei Modulen. Mit dem ersten wird, basierend auf den meteorologischen Vorhersagen des DWD, eine Prognose für den Folgetag erstellt. Das zweite Modul bezieht die online gemessenen Leistungsdaten der repräsentativen Windparks mit ein, um die ursprüngliche Folgetagsprognose zu verbessern und eine sehr genaue Kurzzeitprognose für die nächsten drei bis sechs Stunden zu berechnen. Mit den Ergebnissen der Prognosemodule für die repräsentativen Standorte wird dann über ein Transformationsmodell, dem so genannten Online-Modell, die Gesamteinspeisung in einem größeren Gebiet berechnet. Das Prognoseverfahren hat seine besonderen Vorzüge in der Genauigkeit, den geringen Rechenzeiten und den niedrigen Betriebskosten, da durch die Verwendung des bereits implementierten Online-Modells nur eine geringe Anzahl von Vorhersage- und Messstandorten benötigt wird. Das hier vorgestellte Prognosemodell wurde ursprünglich für die E.ON-Netz GmbH entwickelt und optimiert und ist dort seit Juli 2001 im Einsatz. Es lässt sich jedoch auch leicht an andere Gebiete anpassen. Benötigt werden dazu nur die Messdaten der Leistung ausgewählter repräsentativer Windparks sowie die dazu gehörenden Wettervorhersagen, um die KNN entsprechend zu trainieren.