Melanoma is one of the fastest growing cancers in the world, and can affect patients earlier in life than most other cancers. Therefore, it is imperative to be able to identify patients at high risk for melanoma and enroll them in screening programs to detect the cancer early. Electronic health records collect an enormous amount of data about real-world patient encounters, treatments, and outcomes. This data can be mined to increase our understanding of melanoma as well as build personalized models to predict risk of developing the cancer. Cancer risk models built from structured clinical data are limited in current research, with most studies involving just a few variables from institutional databases or registries. This dissertation presents data processing and machine learning approaches to build melanoma risk models from a large database of de-identified electronic health records. The database contains consistently captured structured data, enabling the extraction of hundreds of thousands of data points each from millions of patient records. Several experiments are performed to build effective models, particularly to predict sentinel lymph node metastasis in known melanoma patients and to predict individual risk of developing melanoma. Data for these models suffer from high dimensionality and class imbalance. Thus, classifiers such as logistic regression, support vector machines, random forest, and XGBoost are combined with advanced modeling techniques such as feature selection and data sampling. Risk factors are evaluated using regression model weights and decision trees, while personalized predictions are provided through random forest decomposition and Shapley additive explanations. Random undersampling on the melanoma risk dataset shows that many majority samples can be removed without a decrease in model performance. To determine how much data is truly needed, we explore learning curve approximation methods on the melanoma data and three publicly-available large-scale biomedical datasets. We apply an inverse power law model as well as introduce a novel semi-supervised curve creation method that utilizes a small amount of labeled data. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2019. / FAU Electronic Theses and Dissertations Collection

Thesis (DTech (Education))--Cape Peninsula University of Technology, Cape Town, 2006 / It is clear that large numbers of tertiary students - particularly those from socially marginalized populations - are not adequately served by traditional models and modalities of education, with the disparate performance rates of majority and minority group members ultimately functioning to reproduce patterns of social inequality. I contend that perhaps even more troubling than these sociological realities is the existence of an ideological underpinning that treats traditional classroom and institutional practices as normative and thus obscures the hegemonic nature of these phenomena. In turn, through the reification of dominant beliefs and habits, the possibilities for more progressive thought and action are inhibited. This thesis addresses the need for educators to consciously interrogate the powerful belief systems that pervade educational discourse and behaviour in order to disrupt oppressive conditions. By introducing the notion of 'conscious practice: I suggest that when educators carefully consider their own hidden cultural biases, seek to more fully understand the perspectives internalized by students, and appreciate the socio-political context in which they operate, they will be in a better position to move towards inclusive and meaningful approaches. Drawing from the central tenets of social justice education, I suggest that lecturers and administrators must intentionally endeavour to promote innovative strategies and structures that enhance learning outcomes for all students and create the conditions for social transformation. The essays contained within this thesis deliver a thorough treatment of this theoretical background as well as provide concrete techniques for bringing these goals to reality.

Orientadores: Siome Klein Goldenstein, Jacques Wainer / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-16T14:41:21Z (GMT). No. of bitstreams: 1 Carvalho_TiagoJosede_M.pdf: 8401323 bytes, checksum: f84374dac5bebf5ea465a7a74ea9b5e4 (MD5) Previous issue date: 2010 / Resumo: O desenvolvimento de métodos computacionais capazes de auxiliar especialistas de diversas áreas na realização de suas tarefas é foco de diversos estudos. Na área da saúde, o diagnóstico precoce de doenças é muito importante para a melhoria da qualidade de vida dos pacientes. Para oftalmologistas que tratam de pacientes com diabetes, um método confiável para a detecção de anomalias em imagens de fundo de olho é importante para um diagnóstico precoce evitando o aparecimento de complicações na retina. Tais complicações podem causar até cegueira. Exsudatos duros é uma das anomalias mais comuns encontradas na retina, sendo sua detecção o foco de vários tipos de abordagens na literatura. Esta dissertação apresenta uma nova e eficiente abordagem para detecção de exsudatos duros em imagens de fundo de olho. Esta abordagem utiliza técnicas de visão computacional e inteligência artificial, como descritores locais, dicionários visuais, agrupamentos e classificação de padrões para detectar exsudatos nas imagens. / Abstract: The computational methods development can helps specialists of several areas in your works is focus of many studies. In health area the premature diagnosis of diseases is very important to improve the patient's life quality. To ophthalmologists who treat patients with diabetics, a reliable method to anomalies detects in eye fundus images is important to a premature diagnosis, avoiding appear of retina complications. Such complications can cause blindness. Hard Exsudates is one of more common anomalies found at retina, being your detection is the focus of many kinds of approaches in literature. This master's thesis presents a new and efficient approach for detection of exsudates at eye fundus images. This approach uses computer vision and artificial inteligence techniques like visiual dictionaries, clustering and pattern recognition to detect hard exsudates in images. / Mestrado / Visão Computacional / Mestre em Ciência da Computação

Deep Neural Networks have been widely applied in many different applications and achieve significant improvement over classical machine learning techniques. However, training a neural network usually requires large amount of data, which is not guaranteed in some applications such as medical image classification. To address this issue, people propose to implement meta learning and ensemble learning techniques to make deep learning trainers more powerful. This thesis focuses on using deep learning equipped with meta learning and ensemble learning to study specific problems. We first propose a new deep learning based method for suggestion mining. The major challenges of suggestion mining include cross domain issue and the issues caused by unstructured and highly imbalanced data structure. To overcome these challenges, we propose to apply Random Multi-model Deep Learning (RMDL) which combines three different deep learning architectures (DNNs, RNNs and CNNs) and automatically selects the optimal hyper parameter to improve the robustness and flexibility of the model. Our experimental results on the SemEval-2019 competition Task 9 data sets demonstrate that our proposed RMDL outperforms most of the existing suggestion mining methods. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2020. / FAU Electronic Theses and Dissertations Collection

Numerous examples arise in fields ranging from mechanics to biology where disappearance of Chaos can be detrimental. Preventing such transient nature of chaos has been proven to be quite challenging. The utility of Reinforcement Learning (RL), which is a specific class of machine learning techniques, in discovering effective control mechanisms in this regard is shown. The autonomous control algorithm is able to prevent the disappearance of chaos in the Lorenz system exhibiting meta-stable chaos, without requiring any a-priori knowledge about the underlying dynamics. The autonomous decisions taken by the RL algorithm are analyzed to understand how the system’s dynamics are impacted. Learning from this analysis, a simple control-law capable of restoring chaotic behavior is formulated. The reverse-engineering approach adopted in this work underlines the immense potential of the techniques used here to discover effective control strategies in complex dynamical systems. The autonomous nature of the learning algorithm makes it applicable to a diverse variety of non-linear systems, and highlights the potential of RLenabled control for regulating other transient-chaos like catastrophic events. / Includes bibliography. / Thesis (M.S.)--Florida Atlantic University, 2020. / FAU Electronic Theses and Dissertations Collection

The success of deep learning has renewed interest in applying neural networks and other machine learning techniques to most fields of data and signal processing, including communications. Advances in architecture and training lead us to consider new modem architectures that allow flexibility in design, continued learning in the field, and improved waveform coding. This dissertation examines neural network architectures and training methods suitable for demodulation in power-limited communication systems, such as those found in wireless sensor networks. Such networks will provide greater connection to the world around us and are expected to contain orders of magnitude more devices than cellular networks. A number of standard and proprietary protocols span this space, with modulations such as frequency-shift-keying (FSK), Gaussian FSK (GFSK), minimum shift keying (MSK), on-off-keying (OOK), and M-ary orthogonal modulation (M-orth). These modulations enable low-cost radio hardware with efficient nonlinear amplification in the transmitter and noncoherent demodulation in the receiver. / Includes bibliography. / Dissertation (Ph.D.)--Florida Atlantic University, 2020. / FAU Electronic Theses and Dissertations Collection

Indiana University-Purdue University Indianapolis (IUPUI) / This dissertation proposes an approach to reduce the cost of manual inspections for as large a number of false positive warnings that are being reported by Static Code Analysis (SCA) tools as much as possible using Machine Learning (ML) techniques. The proposed approach neither assume to use the particular SCA tools nor depends on the specific programming language used to write the target source code or the application. To reduce the number of false positive warnings we first evaluated a number of SCA tools in terms of software engineering metrics using a highlighted synthetic source code named the Juliet test suite. From this evaluation, we concluded that the SCA tools report plenty of false positive warnings that need a manual inspection. Then we generated a number of datasets from the source code that forced the SCA tool to generate either true positive, false positive, or false negative warnings. The datasets, then, were used to train four of ML classifiers in order to classify the collected warnings from the synthetic source code. From the experimental results of the ML classifiers, we observed that the classifier that built using the Random Forests (RF) technique outperformed the rest of the classifiers. Lastly, using this classifier and an instance-based transfer learning technique, we ranked a number of warnings that were aggregated from various open-source software projects. The experimental results show that the proposed approach to reduce the cost of the manual inspection of the false positive warnings outperformed the random ranking algorithm and was highly correlated with the ranked list that the optimal ranking algorithm generated.

To solve real-world problems, machines are required to perceive multiple modalities and fuse the information from them. This thesis studies learning to understand and fuse multimodal information. Existing approaches follow a three-stage learning paradigm. The first stage is to train models for each modality. This process for video understanding models is usually based on supervised training, which is not scalable. Moreover, these modality-specific models are updated rather frequently nowadays with improving single-modality perception abilities. The second stage is crossmodal pretraining, which trains a model to align and fuse multiple modalities based on paired multimodal data, such as video-caption pairs. This process is resource-consuming and expensive. The third stage is to further fine-tune or prompt the resulting model from the second stage towards certain downstream tasks. The key bottleneck of conventional methods lies in the continuous feature representation used for non-textual modalities, which is usually costly to align and fuse with text. In this thesis, we investigate the representation and the fusion based on textual concepts. We propose to map non-textual modalities to textual concepts and then fuse these textual concepts using text models. We systematically study various specific methods of mapping and different architectures for fusion. The proposed methods include an end-to-end video-based text generation model with differentiable tokenization for video and audio concepts, a contrastive-model-based architecture with zero-shot concept extractor, a deep concept injection algorithm enabling language models to solve multimodal tasks without any training, and a distant supervision framework learning concepts in a long temporal span. With our concept representation, we empirically demonstrate that without several orders of magnitude more cost for the crossmodal pretraining stage, our models are able to achieve competitive or even superior performance on downstream tasks such as video question answering, video captioning, text-video retrieval, and audio-video dialogue. We also examine the possible limitations of concept representations such as when the text quality of a dataset is poor. We believe we show a potential path towards upgradable multimodal intelligence, whose components can be easily updated towards new models or new modalities of data.

No description available.

Orientador: Ricardo da Silva Torres / Dissertação (mestrado) - Universidade Estadual de Campinas, Instituto de Computação / Made available in DSpace on 2018-08-17T10:19:12Z (GMT). No. of bitstreams: 1 Panaggio_RicardoLuisZanetti_M.pdf: 1050987 bytes, checksum: f88ede3a681c880be4489f30662ec451 (MD5) Previous issue date: 2010 / Resumo: A diferença entre o conjunto de dados disponíveis e o conjunto dos dados que interessam a um usuário é enorme e, em geral, cresce diariamente, uma vez que o volume de dados produzidos todos os dias só aumenta. Identificar todo o conjunto de dados de interesse de um usuário utilizando mecanismos tradicionais é muito difícil - talvez impossível. Nesse cenário, ferramentas que possam ajudar usuários a identificar itens de interesse, como sistemas de recomendação, têm um grande valor. Esta dissertação apresenta um modelo genérico que pode ser utilizado para a criação de sistemas de recomendação, e sua instanciação utilizando técnicas de agrupamento. Essa dissertação apresenta também a validação desse modelo, a partir de sua implementação e experimentação com dados das bases Movielens e Jester. As principais contribuições são: definição de um modelo de recomendação baseado em grafos, até onde se sabe mais rico e mais genérico que os encontrados na literatura; especificação e implementação de uma arquitetura modular de um sistema de recomendação baseada nesse modelo, com enfoque em técnicas de agrupamento de dados; validação da arquitetura e do modelo de recomendação propostos, comparando eficácia e eficiência de técnicas de agrupamento de dados em sistemas de recomendação / Abstract: The difference between the data available and the set of interesting data to a certain user is enormous and, in general, is becoming greater daily, as the amount of data produced increases. Identifying all the interesting data set using traditional mechanisms is difficult- sometimes impossible. In this scenario, providing tools that can help users on identifying items that are of interest, such as recommendation systems, is of great importance. This dissertation presents a generic model that can be used to create recommender systems, and its instantiation using clustering techniques. It also discusses the validation of this model, by showing results obtained from experiments with data from Movielens and Jester datasets. The main contributions are: a graph-based generic model for recommender systems, which is more generic and richer than the ones found in literature; the specification and implementation of a modular architecture for recommender systems based on that model, focused on clustering techniques; validation of both model and architecture, by comparing efficiency and effectiveness of clustering-based recommender systems / Mestrado / Sistemas de Recuperação da Informação / Mestre em Ciência da Computação