In this thesis, I describe the development of a novel method to selectively modulate neural activity cell-type selectively. Binding of Zolpidem, an allosteric modulator that enhances GABAa receptor function and the inverse agonist β-carboline, require a phenylalanine residue (F77) in the γ subunit.

Las células de la cresta neural (CCN) conforman una población transitoria presente solo en etapas muy tempranas del desarrollo embrionario de vertebrados. Estas células se caracterizan por su multipotencia y capacidad migratoria, y es sabido que defectos en el proceso de migración lleva a severos trastornos congénitos conocidos como neurocristopatias. Las CCN, para adquirir sus propiedades migratorias, sufren un proceso de transición epiteliomesénquima (TEM), similar a lo que ocurre durante el inicio de la metástasis tumoral. Se ha determinado que los microARNs conforman un grupo de reguladores claves de la TEM tumoral. Considerando lo dicho, nos planteamos como objetivo determinar la existencia de una red regulatoria epigenéticamicroARN que desempeñe un papel importante en la migración de las CCN. Es así que, mediante análisis in silico, se encontró que los reguladores claves de la TEM en las CCN, tales como son los genes PHD12 y Snail2, son blancos tentativos del mismo miR-203. Al analizar la expresión de miR-203 mediante hibridación in situ y RT-qPCR se observó que se expresa fuertemente desde estadios muy tempranos en la placa neural y el tubo neural. Sin embargo, su expresión disminuye en el tubo neural dorsal coincidentemente con el inicio de la migración de las CCN. Por otra parte, se observó mediante secuenciación por bisulfito, que la región genómica de miR-203 se encuentra hipermetilada en las CCN pre-migratorias, a diferencia de la hipometilación encontrada en las células del tubo neural ventral y CCN migratorias. Finalmente, la pérdida de función de miR-203, utilizando un vector “esponja” de microARNs, lleva a una migración prematura de las CCN. Estos resultados en conjunto indican que la inhibición de la expresión de miR-203, mediante metilación del ADN, permite el aumento de la expresión de los genes PHD12 y Snail2 los cuales están directamente involucrados en la TEM de las CCN. Los resultados obtenidos en este proyecto pueden tener grandes implicancias que permitirán tener un mayor entendimiento sobre los errores que pueden conducir a un desarrollo anormal, así como también para comprender la metástasis tumoral.

<p>Moral cognition represents a foundational faculty of the human species. Our sense of morality develops beginning at a very young age, and its dysfunction can lead to devastating mental disorders. Given its central importance, it has fittingly garnered the attention of thinkers throughout the ages. For millennia, philosophers have pondered what it is to be right or wrong, good or bad, virtuous or vicious. For centuries, psychologists have elucidated how people acquire and act upon a sense of morality. More recently in the last decade, neuroscientists have embarked on a project to study how morality arises from computations in the brain. However, this latest project has been fragmented: researchers have largely studied various neural components of morality - including emotion, value, and mentalizing - in isolation. This has resulted in an informal and disjointed model for the neural mechanisms of morality. This dissertation is concerned with more formally identifying neural components and their influences on each other in the context of moral cognition.</p><p>In Chapter 2, I study how the component neural networks of moral cognition may be involved in distinct aspects of a single decision by employing a complex clinical decision making task involving the disclosure of conflicts of interest. I show that for a given decision, the magnitude of conflict of interest is tracked by mentalizing networks, while the degree of disclosure-induced behavioral change exhibited by participants is predicted by value networks. In Chapter 3, I move beyond the informal model of morality used in Chapter 2 and previous literature by devising a methodology to identify hierarchical ontologies of neural circuits; such an approach can have implications on further discussions of morality, and more generally, on other aspects of cognitive neuroscience. From this, I present the 50 elemental neural circuits that are fundamental to human cognition and explore how these elements can differentially combine to form emergent neural circuits. In Chapter 4, I use these advances to address morality, uncovering its relevant component neural networks in a data-driven way. I show that neural circuits important in supporting higher-level moral computations include mentalizing and taste. In Chapter 5, I demonstrate an important complexity in a compositional model of morality. I show that one of the components of moral cognition, mentalizing, can paradoxically be influenced by moral judgments themselves. To conclude, I highlight the implications of both theoretical and methodological advances. The hierarchical ontologies of neural circuits may be a profitable framework for the future characterization and study of mental disorders; and to effectively study these circuits, the use of moral judgment and decision-making paradigms will be effective experimental tasks, considering the centrality of moral cognition to who we are, whether in health or illness.</p> / Dissertation

This thesis presents some substantial theoretical analyses and optimal treatments of Kohonen's self-organising map (SOM) algorithm, and explores the practical application potential of the algorithm for vector quantisation, pattern classification, and image processing. It consists of two major parts. In the first part, the SOM algorithm is investigated and analysed from a statistical viewpoint. The proof of its universal convergence for any dimensionality is obtained using a novel and extended form of the Central Limit Theorem. Its feature space is shown to be an approximate multivariate Gaussian process, which will eventually converge and form a mapping, which minimises the mean-square distortion between the feature and input spaces. The diminishing effect of the initial states and implicit effects of the learning rate and neighbourhood function on its convergence and ordering are analysed and discussed. Distinct and meaningful definitions, and associated measures, of its ordering are presented in relation to map's fault-tolerance. The SOM algorithm is further enhanced by incorporating a proposed constraint, or Bayesian modification, in order to achieve optimal vector quantisation or pattern classification. The second part of this thesis addresses the task of unsupervised texture-image segmentation by means of SOM networks and model-based descriptions. A brief review of texture analysis in terms of definitions, perceptions, and approaches is given. Markov random field model-based approaches are discussed in detail. Arising from this a hierarchical self-organised segmentation structure, which consists of a local MRF parameter estimator, a SOM network, and a simple voting layer, is proposed and is shown, by theoretical analysis and practical experiment, to achieve a maximum likelihood or maximum a posteriori segmentation. A fast, simple, but efficient boundary relaxation algorithm is proposed as a post-processor to further refine the resulting segmentation. The class number validation problem in a fully unsupervised segmentation is approached by a classical, simple, and on-line minimum mean-square-error method. Experimental results indicate that this method is very efficient for texture segmentation problems. The thesis concludes with some suggestions for further work on SOM neural networks.

A number of researchers have investigated the application of neural networks to visual recognition, with much of the emphasis placed on exploiting the network's ability to generalise. However, despite the benefits of such an approach it is not at all obvious how networks can be developed which are capable of recognising objects subject to changes in rotation, translation and viewpoint. In this study, we suggest that a possible solution to this problem can be found by studying aspects of visual psychology and in particular, perceptual organisation. For example, it appears that grouping together lines based upon perceptually significant features can facilitate viewpoint independent recognition. The work presented here identifies simple grouping measures based on parallelism and connectivity and shows how it is possible to train multi-layer perceptrons (MLPs) to detect and determine the perceptual significance of any group presented. In this way, it is shown how MLPs which are trained via backpropagation to perform individual grouping tasks, can be brought together into a novel, large scale network capable of determining the perceptual significance of the whole input pattern. Finally the applicability of such significance values for recognition is investigated and results indicate that both the NILP and the Kohonen Feature Map can be trained to recognise simple shapes described in terms of perceptual significances. This study has also provided an opportunity to investigate aspects of the backpropagation algorithm, particularly the ability to generalise. In this study we report the results of various generalisation tests. In applying the backpropagation algorithm to certain problems, we found that there was a deficiency in performance with the standard learning algorithm. An improvement in performance could however, be obtained when suitable modifications were made to the algorithm. The modifications and consequent results are reported here.

Images are forming an increasingly large part of modern communications, bringing the need for efficient and effective compression. Many techniques developed for this purpose include transform coding, vector quantization and neural networks. In this thesis, a new neural network method is used to achieve image compression. This work extends the use of 2-layer neural networks to a combination of cascaded networks with one node in the hidden layer. A redistribution of the gray levels in the training phase is implemented in a random fashion to make the minimization of the mean square error applicable to a broad range of images. The computational complexity of this approach is analyzed in terms of overall number of weights and overall convergence. Image quality is measured objectively, using peak signal-to-noise ratio and subjectively, using perception. The effects of different image contents and compression ratios are assessed. Results show the performance superiority of cascaded neural networks compared to that of fixedarchitecture training paradigms especially at high compression ratios. The proposed new method is implemented in MATLAB. The results obtained, such as compression ratio and computing time of the compressed images, are presented.

No description available.

Master of Science in Engineering - Engineering / Artificial Neural Networks has seen tremendous growth in recent years. It has been applied to various sciences, including applied mathematics, chemistry, physics, and engineering and has also been implemented in various areas of finance. Many researchers have applied them to forecasting of stock prices and other fields of finance. In this study we focus on option pricing. An option is a contract giving the buyer of the contract the right but not the obligation to purchase stock on or before a certain expiration date. Options have become a multi-billion dollar industry in modern times, and there has been a lot of focus on pricing these option contracts. Option pricing data is highly non-linear and its pricing has its basis in stochastic calculus. Since neural networks have excellent non-linear modeling capabilities, it seems obvious to apply neural networks to option pricing. In this thesis, many different methodologies are developed to model the data. The multilayer perceptron and radial basis functions are used in the stand-alone neural networks. Then, the architectures of the stand-alone networks are optimized using particle swarm optimization, which leads to excellent results. Thereafter, a committee of neural networks is investigated. A committee network is an average of a combination of stand-alone neural networks. In contrast to stand-alone networks, a committee network has great generalization capabilities. Many different methods are developed for attaining optimal results from these committee networks. The methods included different forms of weighting the stand-alone networks, a non-linear combination of the committee members using another stand-alone neural network, a two layer committee network where the second layer was used for smoothing the output and a circular committee network. Lastly, genetic algorithm, with the Metropolis-Hastings algorithm, was used to optimize the committee of neural networks. Finally all these methods were analyzed.

A Dissertation submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of the requirements for the degree of Master of Science in Mathematical Statistics,School of Statistics and Actuarial Science. October 2016. / Neural networks (NNs) may be characterised by complex error functions with attributes such as saddle-points, local minima, even-spots and plateaus. This complicates the associated training process in terms of efficiency, convergence and accuracy given that it is done by minimising such complex error functions. This study empirically investigates the performance of two NNs training algorithms which are based on unconstrained and global optimisation theories, i.e. the Resilient propagation (Rprop) and the Conjugate Gradient with Polak-Ribière updates (CGP). It also shows how the network structure plays a role in the training optimisation of NNs. In this regard, various training scenarios are used to classify two protein data, i.e. the Escherichia coli and Yeast data. These training scenarios use varying numbers of hidden nodes and training iterations. The results show that Rprop outperforms CGP. Moreover, it appears that the performance of classifiers varies under various training scenarios. / LG2017

Stereopsis is a par excellence demonstration of the computational power that neural systems can encapsulate. How is the brain capable of swiftly transforming a stream of binocular two-dimensional signals into a cohesive three-dimensional percept? Many brain regions have been implicated in stereoscopic processing, but their roles remain poorly understood. This dissertation focuses on the contributions of primary and dorsomedial visual cortex. Using convolutional neural networks, we found that disparity encoding in primary visual cortex can be explained by shallow, feed-forward networks optimized to extract absolute depth from naturalistic images. These networks develop physiologically plausible receptive fields, and predict neural responses to highly unnatural stimuli commonly used in the laboratory. They do not necessarily relate to our experience of depth, but seem to act as a bottleneck for depth perception. Conversely, neural activity in downstream specialized areas is likely to be a more faithful correlate of depth perception. Using ultra-high field functional magnetic resonance imaging in humans, we revealed systematic and reproducible cortical organization for stereoscopic depth in dorsal visual areas V3A and V3B/KO. Within these regions, depth selectivity was inversely related to depth magnitude — a key characteristic of stereoscopic perception. Finally, we report evidence for a differential contribution of cortical layers in stereoscopic depth perception.