The physical architecture of neural circuits is thought to underlie the computations that give rise to higher order feature sensitivity in the neocortex. Recent technological breakthroughs have allowed the structural and functional investigation of the basic computational units of neural circuits; individual synaptic connections. However, it remains unclear how cortical neurons sample and integrate the thousands of synaptic inputs, supplied by different brain structures, to achieve feature selectivity. Here, I first describe how visual cortical circuits transform the elementary inputs supplied by the periphery into highly diverse, but well-organized, feature representations. By combining and optimizing newly developed techniques to map the functional synaptic connections with defined sources of inputs, I show that the intersection between columnar architecture and dendritic sampling strategies can lead to the selectivity properties of individual neurons: First, in the canonical feedforward circuit, the basal dendrites of a pyramidal neuron utilize unique strategies to sample ON (light increment) and OFF (light decrement) inputs in orientation columns to create the distinctive receptive field structure that is responsible for basic sensitivity to visual spatial location, orientation, spatial frequency, and phase. Second, for long-range horizontal connections, apical dendrites unbiasedly integrate functionally specialized and spatially targeted inputs in different orientation columns, which generates specific axial surround modulation of the receptive field. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2019. / FAU Electronic Theses and Dissertations Collection

No description available.

In this thesis, a novel hybrid Speech Recognition (SR) system called RUST (Recognition Using Syntactical Tree) is developed. RUST combines Artificial Neural Networks (ANN) with a Statistical Knowledge Source (SKS) for a small topic focused database. The hypothesis of this research work was that the inclusion of syntactic knowledge represented in the form of probability of occurrence of phones in words and sentences improves the performance of an ANN-based SR system. The lexicon of the first version of RUST (RUST-I) was developed with 1357 words of which 549 were unique. These words were extracted from three topics (finance, physics and general reading material), and could be expanded or reduced (specialised). The results of experiments carried out on RUST showed that by including basic statistical phonemic/syntactic knowledge with an ANN phone recognisor, the phone recognition rate was increased to 87% and word recognition rate to 78%. The first implementation of RUST was not optimal. Therefore, a second version of RUST (RUST-II) was implemented with an incremental learning algorithm and it has been shown to improve the phone recognition rate to 94%. The introduction of incremental learning to ANN-based speech recognition can be considered as the most innovative feature of this research. In conclusion this work has proved the hypothesis that inclusion of a phonemic syntactic knowledge of probabilistic nature and topic related statistical data using an adaptive phone recognisor based on neural networks has the potential to improve the performance of a speech recognition system. / Doctor of Philosophy (PhD)

A feed forward architecture is suggested that increases the complexity of conventional neural network components through the implementation of a more complex scheme of interconnection. This is done with a view to increasing the range of application of the feed forward paradigm. The uniqueness of this new network design is illustrated by developing an extended taxonomy of accepted published constructs specific and similar to the higher order, product kernel approximations achievable using "parallel paths". Network topologies from this taxonomy are then compared to each other and the architectures containing parallel paths. In attempting this comparison, the context of the term "network topology" is reconsidered. The output of "channels" in these parallel paths are the products of a conventional connection as observed facilitating interconnection between two layers in a multilayered perceptron and the output of a network processing unit, a "control element", that can assume the identity of a number of pre-existing processing paradigms. The inherent property of universal approximation is tested by existence proof and the method found to be inconclusive. In so doing an argument is suggested to indicate that the parametric nature of the functions as determined by conditions upon initialization may only lead to conditional approximations. The property of universal approximation is neither, confirmed or denied. Universal approximation cannot be conclusively determined by the application of Stone Weierstrass Theorem, as adopted from real analysis. This novel implementation requires modifications to component concepts and the training algorithm. The inspiration for these modifications is related back to previously published work that also provides the basis of "proof of concept". By achieving proof of concept the appropriateness of considering network topology without assessing the impact of the method of training on this topology is considered and discussed in some detail. Results of limited testing are discussed with an emphasis on visualising component contributions to the global network output.

The basis of this thesis is the presumption that while neural networks are useful structures that can be used to model complex, highly non-linear systems, current methods of training the neural networks are inadequate in some problem domains. Genetic algorithms have been used to optimise both the weights and architectures of neural networks, but these approaches do not treat the neural network in a sensible manner. In this thesis, I define the basis of computation within a neural network as a single neuron and its associated input connections. Sets of these neurons, stored in a matrix representation, comprise the building blocks that are transferred during one or more epochs of a genetic algorithm. I develop the concept of a Neural Building Block and two new genetic algorithms are created that utilise this concept. The first genetic algorithm utilises the micro neural building block (micro-NBB); a unit consisting of one or more neurons and their input connections. The micro-NBB is a unit that is transmitted through the process of crossover and hence requires the introduction of a new crossover operator. However the micro NBB can not be stored as a reusable component and must exist only as the product of the crossover operator. The macro neural building block (macro-NBB) is utilised in the second genetic algorithm, and encapsulates the idea that fit neural networks contain fit sub-networks, that need to be preserved across multiple epochs. A macro-NBB is a micro-NBB that exists across multiple epochs. Macro-NBBs must exist across multiple epochs, and this necessitates the use of a genetic store, and a new operator to introduce macro-NBBs back into the population at random intervals. Once the theoretical presentation is completed the newly developed genetic algorithms are used to evolve weights for a variety of architectures of neural networks to demonstrate the feasibility of the approach. Comparison of the new genetic algorithm with other approaches is very favourable on two problems: a multiplexer problem and a robot control problem.

Feedforward neural networks have been widely studied and used in many applications in science and engineering. The training of this type of networks is mainly undertaken using the well-known backpropagation based learning algorithms. One major problem with this type of algorithms is the slow training convergence speed, which hinders their applications. In order to improve the training convergence speed of this type of algorithms, many researchers have developed different improvements and enhancements. However, the slow convergence problem has not been fully addressed. This thesis makes several contributions by proposing new backpropagation learning algorithms based on the terminal attractor concept to improve the existing backpropagation learning algorithms such as the gradient descent and Levenberg-Marquardt algorithms. These new algorithms enable fast convergence both at a distance from and in a close range of the ideal weights. In particular, a new fast convergence mechanism is proposed which is based on the fast terminal attractor concept. Comprehensive simulation studies are undertaken to demonstrate the effectiveness of the proposed backpropagataion algorithms with terminal attractors. Finally, three practical application cases of time series forecasting, character recognition and image interpolation are chosen to show the practicality and usefulness of the proposed learning algorithms with comprehensive comparative studies with existing algorithms.

Thesis (Ph. D.)--Ohio State University, 2007. / Title from first page of PDF file. Includes bibliographical references (p. 142-149).

Thesis (Ph. D. in Biomedical Engineering)--Vanderbilt University, May 2007. / Title from title screen. Includes bibliographical references.

Thesis (M.S.)--Brigham Young University. Dept of Computer Science, 2006. / Includes bibliographical references (p. 39-41).

In remotely located forest watersheds, monitoring nitrogen (N) in streams often is not feasible because of the high costs and site inaccessibility. Therefore, modelling tools that can predict N in unmonitored watersheds are urgently needed to support management decisions for these watersheds. Recently, remote sensing (RS) has become a cost-efficient way to evaluate watershed characteristics and obtain model input variables. This study was to develop an artificial neural network (ANN) modelling tool relying solely on public domain climate data and satellite data without ground-based measurements. ANN was successfully applied to simulate N compositions in streams at studied watersheds by using easily accessible input variables, relevant time-lagged inputs and inputs reflecting seasonal cycles. This study was the first effort to take the consideration of vegetation dynamics into N modelling by using RS-derived enhanced vegetation index (EVI) that was capable of describing the differences of vegetation canopy and vegetation dynamics among watersheds. As a further study to demonstrate the applicability of the ANN models to unmonitored watersheds, the calibrated ANN models were used to predict N in other different watersheds (unmonitored watersheds in this perspective) without further calibration. A watershed similarity index was found to show high correlation with the transferability of the models and can potentially guide transferring the trained models into similar unmonitored watersheds. Finally, a framework to incorporate water quantity/quality modelling into forestry management was proposed to demonstrate the application of the developed models to support decision making. The major components of the framework include watershed delineation and classification, database and model development, and scenario-based analysis. The results of scenario analysis can be used to translate vegetation cut into values of EVI that can be fed to the models to predict changes in water quality (e.g. N) in response to harvesting scenarios. The results from this research demonstrated the applicability of ANNs for stream N modelling using easily accessible data, the effectiveness of RS-derived EVI in N model construction, and the transferability of the ANN models. The presented models have high potential to be used to predict N in streams in the real-world and serve forestry management. / Environmental Engineering