Population-wide synchronized rhythmic bursts of electrical activity are present in a variety of neural circuits. The proposed general mechanisms for rhythmogenesis are often attributed to intrinsic and synaptic properties. For example, the recurrent excitation through excitatory synaptic connections determines burst initiation, and the slower kinetics of ionic currents or synaptic depression results in burst termination. In such theories, a slow recovery process is essential for the slow dynamics associated with bursting. This thesis presents a new hypothesis that depends on the connectivity pattern among neurons rather than a slow kinetic process to achieve the network-wide bursting. The thesis begins with an introduction of bursts of electrical activity in a purely excitatory neural network and existing theories explaining this phenomenon. It then covers the small-world approach, which is applied to modify the network structure in the simulation, and the Morris-Lecar (ML) neuron model, which is used as the component cells in the network. Simulation results of the dependence of bursting activity on network connectivity, as well as the inherent network properties explaining this dependence are described. This work shows that the network-wide bursting activity emerges in the small-world network regime but not in the regular or random networks, and this small-world bursting primarily results from the uniform random distribution of long-range connections in the network, as well as the unique dynamics in the ML model. Both attributes foster progressive synchronization in firing activity throughout the network during a burst, and this synchronization may terminate a burst in the absence of an obvious slow recovery process. The thesis concludes with possible future work.

Global climate models (GCM) have been used for nearly two decades now as a tool to investigate and analyze past, present, and future weather and climate. Even though the first several generations of climate models were very simple, today's models are very sophisticated. They use complex parameterization schemes to approximate many nonlinear physical fields. In these models, the resolution and time steps can be set to be as small or as large as desired. In either case, the model generates over 100 atmospheric variables and 20 land surface variables that can be reported daily or monthly. The Community Atmospheric Model Version Two global climate model spends over sixty percent of the time computing shortwave and longwave parameterization schemes. Our goal is to replace its shortwave scheme with empirical methods and show that accuracy of the tropospheric variables is not compromised when using these empirical methods. We found that an autoregressive moving average (ARMA) model can be used to simulate the solar radiation at the top of the model atmosphere. However, the calculated insolation value is only valid for one particular grid point. To simulate the radiation over the entire globe, many ARMA models need to be determined. We also found that large 4-10-10-1 neural networks can be used to simulate the solar radiation to within 2 W m-2. However, much smaller and manageable neural networks can be used to simulate the complete solar insolation term if the neural network only simulates the residual after the annual and diurnal cycles and removed from the field (referred to as the - method). By using the neural network in the - method and by setting the eccentricity term to a constant, we were able to cut the models processing of the solar insolation by at least a factor of four.

The Finite-Difference Time-Domain Method (FDTD) is a very powerful numerical method for the full wave analysis electromagnetic phenomena. Due to its flexibility, it can be used to solve numerous electromagnetic scattering problems on microwave circuits, dielectrics, and electromagnetic absorption in biological tissue at microwave frequencies. However, it needs so much computation time to simulate microwave integral circuits by applying the FDTD method. If the structure we simulated is complicated and we want to obtain accurate frequency domain scattering parameters, the simulation time will be so much longer that the efficiency of simulation will be bad as well. Therefore, in the thesis, we introduce an artificial neural networks (ANN) method called ¡§Finite Impulse Response Neural Networks (FIRNN)¡¨ can speed up the FDTD simulation time. In order to boost the efficiency of the FDTD simulation time by stopping the simulation after a sufficient number of time steps and using FIRNN as a predictor to predict time series signal.

The Finite-Difference Time-Domain Method (FDTD) is a very powerful numerical method for the full wave analysis electromagnetic phenomena. Due to its flexibility, it can be used to solve numerous electromagnetic scattering problems on microwave circuits, dielectrics, and electromagnetic absorption in biological tissue at microwave frequencies. However, it needs so much computation time to simulate microwave integral circuits by applying the FDTD method. If the structure we simulated is complicated and we want to obtain accurate frequency domain scattering parameters, the simulation time will be so much longer that the efficiency of simulation will be bad as well. Therefore, in the thesis, we introduce an artificial neural networks (ANN) method called ¡§Infinite Impulse Response Neural Networks (IIRNN)¡¨ can speed up the FDTD simulation time. In order to boost the efficiency of the FDTD simulation time by stopping the simulation after a sufficient number of time steps and using FIRNN as a predictor to predict time series signal.

Sparse approximation is an important class of optimization problem in signal and image processing applications. This thesis presents an analog solution to this problem, based on the Locally Competitive Algorithm (LCA). A Hopfield-Network-like analog system, operating on sub-threshold currents is proposed as a solution. The results of the circuit components' implementation on the RASP2.8a chip, a Field Programmable Analog Array, are presented.

Thesis (Ph. D.)--University of California, San Diego and San Diego State University, 2004. / Vita. Includes bibliographical references (leaves 109-114).

This work outlines the theoretical underpinnings, method, results, and implications for constructing a discussion list analysis tool that categorizes online, educational discussion list messages into levels of cognitive effort. Purpose The purpose of such a tool is to provide evaluative feedback to instructors who facilitate online learning, to researchers studying computer-supported collaborative learning, and to administrators interested in correlating objective measures of students’ cognitive effort with other measures of student success. This work connects computer–supported collaborative learning, content analysis, and artificial intelligence. Method Broadly, the method employed is a content analysis in which the data from the analysis is modeled using artificial neural network (ANN) software. A group of human coders categorized online discussion list messages, and inter-rater reliability was calculated among them. That reliability figure serves as a measuring stick for determining how well the ANN categorizes the same messages that the group of human coders categorized. Reliability between the ANN model and the group of human coders is compared to the reliability among the group of human coders to determine how well the ANN performs compared to humans. Findings Two experiments were conducted in which artificial neural network (ANN) models were constructed to model the decisions of human coders, and the experiments revealed that the ANN, under noisy, real-life circumstances codes messages with near-human accuracy. From experiment one, the reliability between the ANN model and the group of human coders, using Cohen’s kappa, is 0.519 while the human reliability values range from 0.494 to 0.742 (M=0.6). Improvements were made to the human content analysis with the goal of improving the reliability among coders. After these improvements were made, the humans coded messages with a kappa agreement ranging from 0.816 to 0.879 (M=0.848), and the kappa agreement between the ANN model and the group of human coders is 0.70.

Wheel slip is a common problem in planetary rover exploration tasks. During the current Mars Exploration Rover (MER) mission, the Spirit rover almost became trapped on a dune because of wheel slip. As rover missions on harsh terrains expand in scope, mission success will depend not only on rover safety, but also alacrity in task completion. Speed combined with exploration of varied and difficult terrains, the risk of slip increases dramatically. We first characterize slip performance of a rover on harsh terrains by implementing a novel High Fidelity Traversability Analysis (HFTA) algorithm in order to provide slip prediction and detection capabilities to a planetary rover. The algorithm, utilizing path and energy cost functions in conjunction with simulated navigation, allows a rover to select the best path through any given terrain by predicting high slip paths. Integrated software allows the rover to then accurately follow a designated path while compensating for slippage, and reach intended goals independent of the terrain over which it is traversing. The algorithm was verified using ROAMS, a high fidelity simulation package, at 3.5x real time speed. We propose an adaptive path following algorithm as well as a human-trained neural network to traverse multiple harsh terrains using slip as an advantage. On a near-real-time system, and at rover speeds 15 times the current average speed of the Mars Exploration Rovers, we show that the adaptive algorithm traverses paths in less time than a standard path follower. We also train a standard back-propagation neural network, using human and path following data from a near-real-time system. The neural network demonstrates it ability to traverse new paths on multiple terrains and utilize slip to minimize time and path error.

Thesis (Ph. D.)--University of Washington, 2007. / Vita. Includes bibliographical references (p. 175-193).

Orientador: Fernando Antônio Campos Gomide / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-22T00:50:20Z (GMT). No. of bitstreams: 1 Bordignon_FernandoLuis_M.pdf: 1666872 bytes, checksum: 4d838dfb4ec418698d9ecd3b74e7c981 (MD5) Previous issue date: 2013 / Resumo: Sistemas evolutivos são sistemas com alto nível de adaptação capazes de modificar simultaneamente suas estruturas e parâmetros a partir de um fluxo de dados, recursivamente. Aprendizagem a partir de fluxos de dados é um problema contemporâneo e difícil devido à taxa de aumento da dimensão, tamanho e disponibilidade temporal de dados, criando dificuldades para métodos tradicionais de aprendizado. Esta dissertação, além de apresentar uma revisão da literatura de sistemas evolutivos e redes neurais fuzzy, aborda uma estrutura e introduz um método de aprendizagem evolutivo para treinar redes neurais híbridas baseadas em uninormas, usando conceitos de aprendizado extremo. Neurônios baseados em uninormas fundamentados nas normas e conormas triangulares generalizam neurônios fuzzy. Uninormas trazem flexibilidade e generalidade a modelos neurais fuzzy, pois elas podem se comportar como normas triangulares, conormas triangulares, ou de forma intermediária por meio do ajuste de elementos identidade. Este recurso adiciona uma forma de plasticidade em modelos de redes neurais. Um método de agrupamento recursivo para granularizar o espaço de entrada e um esquema baseado no aprendizado extremo compõem um algoritmo para treinar a rede neural. _E provado que uma versão estática da rede neural fuzzy baseada em uninormas aproxima funções contínuas em domínios compactos, ou seja, _e um aproximador universal. Postula-se, e experimentos computacionais endossam, que a rede neural fuzzy evolutiva compartilha capacidade de aproximação equivalente, ou melhor, em ambientes dinâmicos, do que as suas equivalentes estáticas / Abstract: Evolving systems are highly adaptive systems able to simultaneously modify their structures and parameters from a stream of data, online. Learning from data streams is a contemporary and challenging issue due to the increasing rate of the size and temporal availability of data, turning the application of traditional learning methods limited. This dissertation, in addition to reviewing the literature of evolving systems and neuro fuzzy networks, addresses a structure and introduces an evolving learning approach to train uninorm-based hybrid neural networks using extreme learning concepts. Uninorm-based neurons, rooted in triangular norms and conorms, generalize fuzzy neurons. Uninorms bring flexibility and generality to fuzzy neuron models as they can behave like triangular norms, triangular conorms, or in between by adjusting identity elements. This feature adds a form of plasticity in neural network modeling. An incremental clustering method is used to granulate the input space, and a scheme based on extreme learning is developed to train the neural network. It is proved that a static version of the uninorm-based neuro fuzzy network approximate continuous functions in compact domains, i.e. it is a universal approximator. It is postulated and computational experiments endorse, that the evolving neuro fuzzy network share equivalent or better approximation capability in dynamic environments than their static counterparts / Mestrado / Engenharia de Computação / Mestre em Engenharia Elétrica