<p> Medical informatics plays an important role in precision medicine, delivering the right information to the right person, at the right time. With the introduction and widespread adoption of electronic medical records, in the United States and world-wide, there is now a tremendous amount of health data available for analysis.</p><p> Electronic record phenotyping refers to the task of determining, from an electronic medical record entry, a concise descriptor of the patient, comprising of their medical history, current problems, presentation, etc. In inferring such a phenotype descriptor from the record, a computer, in a sense, "understands'' the relevant parts of the record. These phenotypes can then be used in downstream applications such as cohort selection for retrospective studies, real-time clinical decision support, contextual displays, intelligent search, and precise alerting mechanisms.</p><p> We are faced with three main challenges:</p><p> First, the unstructured and incomplete nature of the data recorded in the electronic medical records requires special attention. Relevant information can be missing or written in an obscure way that the computer does not understand. </p><p> Second, the scale of the data makes it important to develop efficient methods at all steps of the machine learning pipeline, including data collection and labeling, model learning and inference.</p><p> Third, large parts of medicine are well understood by health professionals. How do we combine the expert knowledge of specialists with the statistical insights from the electronic medical record?</p><p> Probabilistic graphical models such as Bayesian networks provide a useful abstraction for quantifying uncertainty and describing complex dependencies in data. Although significant progress has been made over the last decade on approximate inference algorithms and structure learning from complete data, learning models with incomplete data remains one of machine learning&rsquo;s most challenging problems. How can we model the effects of latent variables that are not directly observed?</p><p> The first part of the thesis presents two different structural conditions under which learning with latent variables is computationally tractable. The first is the "anchored'' condition, where every latent variable has at least one child that is not shared by any other parent. The second is the "singly-coupled'' condition, where every latent variable is connected to at least three children that satisfy conditional independence (possibly after transforming the data). </p><p> Variables that satisfy these conditions can be specified by an expert without requiring that the entire structure or its parameters be specified, allowing for effective use of human expertise and making room for statistical learning to do some of the heavy lifting. For both the anchored and singly-coupled conditions, practical algorithms are presented.</p><p> The second part of the thesis describes real-life applications using the anchored condition for electronic phenotyping. A human-in-the-loop learning system and a functioning emergency informatics system for real-time extraction of important clinical variables are described and evaluated.</p><p> The algorithms and discussion presented here were developed for the purpose of improving healthcare, but are much more widely applicable, dealing with the very basic questions of identifiability and learning models with latent variables - a problem that lies at the very heart of the natural and social sciences.</p>

<p> Introduced in 2006, Deep Learning has made large strides in both supervised an unsupervised learning. The abilities of Deep Learning have been shown to beat both generic and highly specialized classification and clustering techniques with little change to the underlying concept of a multi-layer perceptron. Though this has caused a resurgence of interest in neural networks, many of the drawbacks and pitfalls of such systems have yet to be addressed after nearly 30 years: speed of training, local minima and manual testing of hyper-parameters.</p><p> In this thesis we propose using an evolutionary technique in order to work toward solving these issues and increase the overall quality and abilities of Deep Learning Networks. In the evolution of a population of autoencoders for input reconstruction, we are able to abstract multiple features for each autoencoder in the form of hidden nodes, scoring the autoencoders based on their ability to reconstruct their input, and finally selecting autoencoders for crossover and mutation with hidden nodes as the chromosome. In this way we are able to not only quickly find optimal abstracted feature sets but also optimize the structure of the autoencoder to match the features being selected. This also allows us to experiment with different training methods in respect to data partitioning and selection, reducing overall training time drastically for large and complex datasets. This proposed method allows even large datasets to be trained quickly and efficiently with little manual parameter choice required by the user, leading to faster, more accurate creation of Deep Learning Networks.</p>

No description available.

No description available.

<p> Feature engineering is a process that augments the feature vector of a machine learning model with calculated values that are designed to enhance the accuracy of a model's predictions. Research has shown that the accuracy of models such as deep neural networks, support vector machines, and tree/forest-based algorithms sometimes benefit from feature engineering. Expressions that combine one or more of the original features usually create these engineered features. The choice of the exact structure of an engineered feature is dependent on the type of machine learning model in use. Previous research demonstrated that various model families benefit from different types of engineered feature. Random forests, gradient-boosting machines, or other tree-based models might not see the same accuracy gain that an engineered feature allowed neural networks, generalized linear models, or other dot-product based models to achieve on the same data set. </p><p> This dissertation presents a genetic programming-based algorithm that automatically engineers features that increase the accuracy of deep neural networks for some data sets. For a genetic programming algorithm to be effective, it must prioritize the search space and efficiently evaluate what it finds. This dissertation algorithm faced a potential search space composed of all possible mathematical combinations of the original feature vector. Five experiments were designed to guide the search process to efficiently evolve good engineered features. The result of this dissertation is an automated feature engineering (AFE) algorithm that is computationally efficient, even though a neural network is used to evaluate each candidate feature. This approach gave the algorithm a greater opportunity to specifically target deep neural networks in its search for engineered features that improve accuracy. Finally, a sixth experiment empirically demonstrated the degree to which this algorithm improved the accuracy of neural networks on data sets augmented by the algorithm's engineered features. </p>

The practical research undertaken for this thesis - the building of interactive and non-interactive environments for performance - posits a radical recasting of the performing body in physical and digital space. The choreographic and thematic context of the performance work has forced us', as makers, to ask questions about the nature of digital interactivity which in turn feeds the work theoretically, technically and thematically. A computer views (and attempts to interpret) motion information through a video camera, and, by way of a scripting language, converts that information into MIDI' data. As the research has developed, our company has been able to design environments which respond sensitivelyto particular artistic / performance demands. I propose to show in this research that is it possible to design an interactive system that is part of a phenomenological performance space, a mechanical system with an ontological heart. This represents a significant shift in thinking from existing systems, is at the heart of the research developments and is what I consider to be one of the primary outcomes of this research, outcomes that are original and contribute to the body of knowledge in this area. The phenomenal system allows me to use technology in a poetic way, where the poetic aesthetic is dominant - it responds to the phenomenal dancer, rather than merely to the 'physico-chemical' (Merleau-Ponty 1964 pp. 10-I I) dancer. Other artists whose work attempts phenomenological approaches to working with technology and the human body are referenced throughout the writing.

<p> The technological advancement and the escalating interest in personalized medicine has resulted in increased ordinal classification problems. The most commonly used performance metrics for evaluating the effectiveness of a multi-class ordinal classifier include; predictive accuracy, Kendall's tau-b rank correlation, and the average mean absolute error (AMAE). These metrics are beneficial in the quest to classify multi-class ordinal data, but no single performance metric incorporates the misclassification cost. Recently, distance, which finds the optimal trade-off between the predictive accuracy and the misclassification cost was proposed as a cost-sensitive performance metric for ordinal data. This thesis proposes the criteria for variable selection and methods that accounts for minimum distance and improved accuracy, thereby providing a platform for a more comprehensive and comparative analysis of multiple ordinal classifiers. The strengths of our methodology are demonstrated through real data analysis of a colon cancer data set.</p>

This thesis brings together two strands of neural networks research - weightless systems and statistical learning theory - in an attempt to understand better the learning and generalisation abilities of a class of pattern classifying machines. The machines under consideration are n-tuple classifiers. While their analysis falls outside the domain of more widespread neural networks methods the method has found considerable application since its first publication in 1959. The larger class of learning systems to which the n-tuple classifier belongs is known as the set of weightless or RAM-based systems, because of the fact that they store all their modifiable information in the nodes rather than as weights on the connections. The analytical tools used are those of statistical learning theory. Learning methods and machines are considered in terms of a formal learning problem which allows the precise definition of terms such as learning and generalisation (in this context). Results relating the empirical error of the machine on the training set, the number of training examples and the complexity of the machine (as measured by the Vapnik- Chervonenkis dimension) to the generalisation error are derived. In the thesis this theoretical framework is applied for the first time to weightless systems in general and to n-tuple classifiers in particular. Novel theoretical results are used to inspire the design of related learning machines and empirical tests are used to assess the power of these new machines. Also data-independent theoretical results are compared with data-dependent results to explain the apparent anomalies in the n-tuple classifier's behaviour. The thesis takes an original approach to the study of weightless networks, and one which gives new insights into their strengths as learning machines. It also allows a new family of learning machines to be introduced and a method for improving generalisation to be applied.

The use of inductive learning to extract general rules from examples would be a promising way to overcome the knowledge acquisition bottleneck. Over the last decade, many such techniques have been proposed. None of these have proved to be the efficient, general rule-extractors for complex real-world applications. Recent research has indicated that some kinds of hybrid-learning techniques which integrate two or more learning strategies outperform single learning techniques. In designing such a hybrid-learning method, neural network learning can be expected to be a good partner because it is tolerant for noisy data and is very flexible for approximate data. / This dissertation proposes another such method--a rule extraction method using an artificial neural network (ANN) that is trained by destructive learning. Unlike other published methods, the method proposed here takes advantage of the smart (pruned) network which contains more exact knowledge regarding the problem domain (environment). The method consists of three phases: training, pruning, and rule-extracting. The training phase is concerned with ANN learning, using a general backpropagation (BP) learning algorithm. In the pruning phase, redundant hidden units and links are deleted from a trained network, and then, the link weights remaining in the network are retrained to obtain near-saturated outputs from hidden units. The rule extraction algorithm uses the pruned network to extract rules. / The proposed method is evaluated empirically on three application domains--the MONK's problems, the IRIS-classification data set, and the thyroid-disease diagnosis data set--and its performance is compared with that of other classification and/or machine learning methods. It is shown that for discrete samples, the proposed method outperforms others, while for continuous samples it can beat most other methods with which it is compared. The classifying accuracy of the proposed method is higher than that of either backpropagation learning or the pruned network on which it is based. / Source: Dissertation Abstracts International, Volume: 55-04, Section: B, page: 1526. / Major Professor: R. C. Lacher. / Thesis (Ph.D.)--The Florida State University, 1994.

This dissertation presents Modified Election (or ME) methodology and shows how it may be used to describe the beliefs a human expert would form regarding the answer to a given question, based on the available evidence. For example, the methodology could be used to describe the beliefs a heart specialist would form, regarding the question whether a patient should be put on a low fat, low cholesterol diet, based on whether the patient is overweight, has a family history of heart problems, etc. ME methodology employs statistical methods used to interpret random samples, as well as the concept of a "Modified Election" which is developed in this dissertation. In ME methodology, the numbers of "votes" for the possible outcomes in a modified election are used to weight the different pieces of evidence which might affect an expert's beliefs. / Two other popular formalisms for describing beliefs are Bayesian theory and Dempster/Shafer theory. Certain problematic aspects of these two formalisms which motivated ME methodology are discussed. It is then shown how ME methodology overcomes these problems. ME methodology may be used as the basis for the design of expert systems. An expert system is presented which illustrates how to do this. / Source: Dissertation Abstracts International, Volume: 54-04, Section: B, page: 2068. / Major Professor: Daniel G. Schwartz. / Thesis (Ph.D.)--The Florida State University, 1993.