Please read the abstract in the section 00front of this document / Dissertation (MSc)--University of Pretoria, 2005. / Computer Science / unrestricted

Reinforcement learning algorithms comprise a class of learning algorithms for neural networks. Reinforcement learning is distinguished from other classes by the type of problems that it is intended to solve. It is used for learning input-output mappings where the desired outputs are not known and only a scalar reinforcement value is available. Primary Reinforcement Learning (PRL) is a core component of the most actively researched form of reinforcement learning. The issues surrounding the convergence characteristics of PRL are considered in this thesis. There have been no convergence proofs for any kind of networks learning under PRL. A convergence theorem is proved in this thesis, showing that under some conditions, a particular reinforcement learning algorithm, the A[formula omitted] algorithm, will train a single-layer network correctly. The theorem is demonstrated with a series of simulations. A new PRL algorithm is proposed to deal with the training of multiple layer, binary output networks with continuous inputs. This is a more difficult learning problem than with binary inputs. The new algorithm is shown to be able to successfully train a network with multiple outputs when the environment conforms to the conditions of the convergence theorem for a single-layer network. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

Neuropsychiatric disorders severely impact quality of life in millions of patients, contributing more Disease Affected Life Years (DALYs) than cancer or cardiovascular disease. The human brain is a complex system of 100 billion neurons connected by 100 trillion synapses, and human studies of neural disease focus on network-level circuit activity changes, rather than on cellular mechanisms. To probe for neural dynamics on the cellular level, animal models such as the nematode C. elegans have been used to investigate the biochemical and genetic factors contributing to neurological disease. C. elegans are ideal for neurophysiological studies due to their small nervous system, neurochemical homology to humans, and compatibility with non-invasive neural imaging. To better study the cellular mechanisms contributing to neurological disease, we developed automated analysis methods for characterizing the behaviors and associated neural activity during sleep and learning in C. elegans: two neural functions that involve a high degree of behavioral and neural plasticity. We developed two methods to study previously uncharacterized spontaneous adult sleep in C. elegans. A large microfluidic device facilitates population-wide assessment of long-term sleep behavior over 12 hours including effects of fluid flow, oxygen, feeding, odors, and genetic perturbations. Smaller devices allow simultaneous recording of sleep behavior and neuronal activity. Since the onset of adult sleep is stochastically timed, we developed a closed-loop sleep detection system that delivers chemical stimuli to individual animals during sleep and awake states to assess state-dependent changes to neural responses. Sleep increased the arousal threshold to aversive chemical stimulation, yet sensory neuron (ASH) and first-layer interneuron (AIB) responses were unchanged. This localizes adult sleep-dependent neuromodulation within interneurons presynaptic to the AVA premotor interneurons, rather than afferent sensory circuits. Traditionally, the study of learning in C. elegans observes taxis on agar plates which present variable environmental conditions that can lead to a reduction in test-to-test reproducibility. We also translated the butanone enhancement learning assay such that animals can be trained and tested all within the controlled environment of a microfluidic device. Using this system, we demonstrated that C. elegans are capable of associative learning by observing stimulus evoked behavioral responses, rather than taxis. This system allows for more reproducible results and can be used to seamlessly study stimulus-evoked neural plasticity associated with learning. Together, these systems provide platforms for studying the connections between behavioral plasticity and neural circuit modulation in sleep and learning. We can use these systems to further our understanding of the mechanisms underlying neural regulation, function, and disorder using human disease models in C. elegans.

Gliomas are an aggressive class of brain tumors that are associated with a better prognosis at a lower grade level. Effective differentiation and classification are imperative for early treatment. MRI scans are a popular medical imaging modality to detect and diagnosis brain tumors due to its capability to non-invasively highlight the tumor region. With the rise of deep learning, researchers have used convolution neural networks for classification purposes in this domain, specifically pre-trained networks to reduce computational costs. However, with various MRI modalities, MRI machines, and poor image scan quality cause different network structures to have different performance metrics. Each pre-trained network is designed with a different structure that allows robust results given specific problem conditions. This thesis aims to cover the gap in the literature to compare the performance of popular pre-trained networks on a controlled dataset that is different than the network trained domain. / Includes bibliography. / Thesis (M.S.)--Florida Atlantic University, 2020. / FAU Electronic Theses and Dissertations Collection

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第22639号 / 医博第4622号 / 新制||医||1044(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 林 康紀, 教授 伊佐 正, 教授 斎藤 通紀 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

No description available.

Optimism refers to peoples’ general tendency to anticipate good outcomes in areas that are important to them. Numerous studies have shown that optimism is significantly correlated with improved physical and mental health. Optimism can come to an overly optimistic degree, called optimism bias. People generally expect better outcomes and fewer negative events to happen for themselves in the future compared to the average person. There are two sides to this: being optimistically biased might lead to risky behavior, but it might also ease people's worries about the future. To have a consistently negative view is suggested to correlate with depressive symptoms and worsened health. The aim of this thesis is to investigate the neural correlates and functional markers of optimism and optimism bias. Optimism is suggested to correlate with gray-matter volume in the thalamus, orbitofrontal cortex (OFC), and bilateral putamen. The inferior frontal gyrus (IFG) and the rostral anterior cingulate cortex (rACC) have a crucial role in dismissing undesirable information and self referential processing. Research regarding this issue might be beneficial for further understanding of the connection between optimism and well-being.

<p>Through the computing advance and GPU (Graphics Processing Unit) availability for math calculation, the deep learning field becomes more popular and prevalent. Object detection with deep learning, which is the part of image processing, plays an important role in automatic vehicle drive and computer vision. Object detection includes object localization and object classification. Object localization involves that the computer looks through the image and gives the correct coordinates to localize the object. Object classification is that the computer classification targets into different categories. The traditional image object detection pipeline idea is from Fast/Faster R-CNN [32] [58]. The region proposal network generates the contained objects areas and put them into classifier. The first step is the object localization while the second step is the object classification. The time cost for this pipeline function is not efficient. Aiming to address this problem, You Only Look Once (YOLO) [4] network is born. YOLO is the single neural network end-to-end pipeline with the image processing speed being 45 frames per second in real time for network prediction. In this thesis, the convolution neural networks are introduced, including the state of art convolutional neural networks in recently years. YOLO implementation details are illustrated step by step. We adopt the YOLO network for our applications since the YOLO network has the faster convergence rate in training and provides high accuracy and it is the end to end architecture, which makes networks easy to optimize and train. </p>

Neuropsychiatric disorders severely impact quality of life in millions of patients, contributing more Disease Affected Life Years (DALYs) than cancer or cardiovascular disease. The human brain is a complex system of 100 billion neurons connected by 100 trillion synapses, and human studies of neural disease focus on network-level circuit activity changes, rather than on cellular mechanisms. To probe for neural dynamics on the cellular level, animal models such as the nematode C. elegans have been used to investigate the biochemical and genetic factors contributing to neurological disease. C. elegans are ideal for neurophysiological studies due to their small nervous system, neurochemical homology to humans, and compatibility with non-invasive neural imaging. To better study the cellular mechanisms contributing to neurological disease, we developed automated analysis methods for characterizing the behaviors and associated neural activity during sleep and learning in C. elegans: two neural functions that involve a high degree of behavioral and neural plasticity. We developed two methods to study previously uncharacterized spontaneous adult sleep in C. elegans. A large microfluidic device facilitates population-wide assessment of long-term sleep behavior over 12 hours including effects of fluid flow, oxygen, feeding, odors, and genetic perturbations. Smaller devices allow simultaneous recording of sleep behavior and neuronal activity. Since the onset of adult sleep is stochastically timed, we developed a closed-loop sleep detection system that delivers chemical stimuli to individual animals during sleep and awake states to assess state-dependent changes to neural responses. Sleep increased the arousal threshold to aversive chemical stimulation, yet sensory neuron (ASH) and first-layer interneuron (AIB) responses were unchanged. This localizes adult sleep-dependent neuromodulation within interneurons presynaptic to the AVA premotor interneurons, rather than afferent sensory circuits. Traditionally, the study of learning in C. elegans observes taxis on agar plates which present variable environmental conditions that can lead to a reduction in test-to-test reproducibility. We also translated the butanone enhancement learning assay such that animals can be trained and tested all within the controlled environment of a microfluidic device. Using this system, we demonstrated that C. elegans are capable of associative learning by observing stimulus evoked behavioral responses, rather than taxis. This system allows for more reproducible results and can be used to seamlessly study stimulus-evoked neural plasticity associated with learning. Together, these systems provide platforms for studying the connections between behavioral plasticity and neural circuit modulation in sleep and learning. We can use these systems to further our understanding of the mechanisms underlying neural regulation, function, and disorder using human disease models in C. elegans.

[pt] Nesta dissertação é desenvolvido um modelo de previsão de energia elétrica de curto prazo (previsão mensal) para o sistema elétrico no Brasil, em especial para as concessionárias dos sistemas interligados, através de um modelo de Redes Neurais que emprega um algoritmo de otimização camada a camada. O objetivo principal deste trabalho consiste em demonstrar que bons resultados preditivos podem ser alcançados com a utilização desse algoritmo para séries de energia elétrica e que esse método poderia fazer parte dos métodos de previsão que compõem o Sistema de Previsão de Carga (PREVCAR) do Operador Nacional do Sistema (ONS) a saber: modelo de Holt & Winters, modelo de Box & Jenkins, modelo de redes Neurais (backpropagation) e modelo de Lógica Fuzzy. / [en] It is developed in this essay a short forecast electric energy model (monthly forecast) to the electric system in Brazil, particularly to interconnected systems utilities, through a neural network model, which employs a layer by layer improvement algorithm. The aim of this proposition consists in demonstrating that good forecast results can be reached with the use this algorithm to electric energy series and that this method could be part of the forecast methods, wich compose the Load Forecasting System (PREVCAR) of National System model (backpropagation) and Fuzzy logic model.