This thesis consists of results relating to the theoretical and computational advances in modeling the Normal-Normal hierarchical model. / Statistics

Thesis (M.S.)--University of Michigan, 1951.

This thesis illustrates how I hear processes of expansion organizing musical materials in the First String Quartet. By employing a flexible approach to expansion and developing models of wedge and additive expansions beyond the bounds of specific voice-leading or rhythmic augmentation procedures, expansion processes can be understood in each of the varied episodes of the quartet. Gubaidulina’s use of expansion processes, embodied organically in pitch, rhythm, form, and physical space, unifies the episodic materials of the First String Quartet and provides an inevitable conclusion to the work’s loose narrative.

Patients with pes planovalgus deformity often have coexisting spring ligament pathology. A primary repair of the ligament may fail during weightbearing due to chronic degeneration of the ligamentous tissue. Augmentation with a suture tape has been suggested to strengthen the repair. Limited data exist regarding flatfoot reconstruction with augmented spring ligament repair using a suture tape. This is a review of 57 consecutive patients who had flatfoot reconstruction with concomitant spring ligament augmented repair between July 2014 and August 2017. Weightbearing radiographic parameters were obtained preoperatively and compared to radiographs at an average time of 62 ± 46.5 (range 20-220) weeks postoperative. Significant improvements were seen in the radiographic parameters evaluated. Five patients had subsequent operations including one deep infection, 2 hardware removals remote to the spring ligament augmentation, 1 ankle arthrodesis, and 1 triple arthrodesis. Concomitant spring ligament repair augmented with a suture tape was a safe procedure that contributed to radiographic correction in a consecutive series of 57 patients undergoing flatfoot deformity correction.

In order to treat cancer effectively, medical practitioners must predict pathological stages accurately, and machine learning methods can be employed to make such predictions. However, biomedical datasets, including genomic datasets, often have disproportionately more samples from people of European ancestry than people of other ethnic or racial groups, which can cause machine learning methods to perform better on the European samples than on the people of the under-represented groups. Data augmentation can be employed as a potential solution in order to artificially increase the number of samples from people of under-represented racial groups, and can in turn improve pathological stage predictions for future patients from such under-represented groups. Genomic data augmentation has been explored previously, for example using a Generative Adversarial Network, but to the best of our knowledge, the use of the variational autoencoder for the purpose of genomic data augmentation remains largely unexplored. Here we utilize a geometry-based variational autoencoder that models the latent space as a Riemannian manifold so that samples can be generated without the use of a prior distribution to show that the variational autoencoder can indeed be used to reliably augment genomic data. Using TCGA prostate cancer genotype data, we show that our VAE-generated data can improve pathological stage predictions on a test set of European samples. Because we only had European samples that were labeled in terms of pathological stage, we were not able to validate the African generated samples in this way, but we still attempt to show how such samples may be realistic. / Computer and Information Science

Paraphrase sparsity is an issue that complicates the training process of question answering systems: syntactically diverse but semantically equivalent sentences can have significant disparities in predicted output probabilities. We propose a method for generating an augmented paraphrase corpus for the visual question answering system to make it more robust to paraphrases. This corpus is generated by concatenating two sequence to sequence models. In order to generate diverse paraphrases, we sample the neural network using diverse beam search. We evaluate the results on the standard VQA validation set. Our approach results in a significantly expanded training dataset and vocabulary size, but has slightly worse performance when tested on the validation split. Although not as fruitful as we had hoped, our work highlights additional avenues for investigation into selecting more optimal model parameters and the development of a more sophisticated paraphrase filtering algorithm. The primary contribution of this work is the demonstration that decent paraphrases can be generated from sequence to sequence models and the development of a pipeline for developing an augmented dataset. / Master of Science

Credential theft is a network intrusion vector that subverts traditional defenses of a campus network, with a malicious login being the act of an attacker using those stolen credentials to access the target network. Historically, this approach is simple for an attacker to conduct and hard for a defender to detect. Alternative mitigation strategies require an in depth view of the network hosts, an untenable proposition in a campus network. We introduce a method of spatial augmentation of login events, creating a user and source IP trajectory for each event. These location mappings, built using user wireless activity and network state information, provide features needed for login classification. From this, we design and build a real time data collection, augmentation, and classification system for generating alerts on malicious events. With a relational database for data processing and a trained weighted random forests ensemble classifier, generated alerts are both timely and few enough to allow human analyst review of all generated events. We evaluate this design for three levels of attacker ability with a defined threat model. We evaluate our approach with a proof of concept system on weeks of live data collected from the Virginia Tech campus, under an IRB approved research protocol. / Master of Science / For a computer network, a common mode of access is a login; the entering of a valid username and password for authentication. Attackers use a variety of methods to steal user login credentials and several of these approaches are unnoticeable by network defenders. Providing further complications, a higher educational campus network, such as Virginia Tech, inherently has less information about the state of the network, since students and teachers bring their privately owned devices. To prevent this attack method, we determine the class, authorized or unauthorized, of login events using data that can be consistently provided by a campus network. After classification, alerts are generated for security analysts, helping to further defend the network. Spatial augmentation is a process we introduce to allow login classification with machine learning algorithms. For every login event at the campus, a history of user locations and source event locations can be provided, using data collected from the campus network infrastructure. Location data provides stronger classification of login events, since studies show attackers inherently have a physical distance between the normal user of an account when performing an unauthorized login. For evaluation, we build a system to augment and classify login events, while limiting the number of false alerts to a useable level.

Breast and penis augmentations are century old processes of body modification continuing in development and practice today. This Masters thesis is a content analysis of breast and penis augmentation print and internet advertisements to explore one facet of augmentation discourse presented in public space. Relevant theoretical literature includes fetish discourse and medical discourse as existing frameworks that conceptualize augmentation predominantly as a process of body fragmentation. After reviewing this literature, I expand to blend together perspectives from three body theorists, Maurice Merleau-Ponty, Michel Foucault and Judith Butler, as an alternate framework for embodiment. I then use content analysis to examine the data collected from 21 print sources and 27 internet sources. The data indicates fetish discourse, focusing on body fragmentation, is the dominant content in breast and penis augmentation advertisements; however, I argue in the conclusion that incorporating elements of embodiment into fetish discourse is a better perspective for future research.

Le contenu de cette thèse est divisé en deux parties. La première partie de cette thèse porte sur l'étude d'approches heuristiques pour approximer des fronts de Pareto. Nous proposons un nouvel algorithme de recherche locale pour résoudre des problèmes d'optimisation combinatoire. Cette technique est intégrée dans un modèle opérationnel générique où l'algorithme évolue vers de nouvelles solutions formées en combinant des solutions trouvées dans les étapes précédentes. Cette méthode améliore les algorithmes de recherche locale existants pour résoudre le problème d'assignation quadratique bi- et tri-objectifs.La seconde partie se focalise sur les algorithmes d'ordonnancement dans un contexte non-préemptif. Plus précisément, nous étudions le problème de la minimisation du stretch maximum sur une seule machine pour une exécution online. Nous présentons des résultats positifs et négatifs, puis nous donnons une solution optimale semi-online. Nous étudions ensuite le problème de minimisation du stretch sur une seule machinedans le modèle récent de la réjection. Nous montrons qu'il existe un rapport d'approximation en O(1) pour minimiser le stretch moyen. Nous montrons également qu'il existe un résultat identique pour la minimisation du flot moyen sur une machine. Enfin, nous étudions le problème de la minimisation du somme des flots pondérés dans un contexte online. / The content of this thesis is divided into two parts. The first part of the thesis deals with the study of heuristic based approaches for the approximation Pareto fronts. We propose a new Double Archive Pareto local search algorithm for solving multi-objective combinatorial optimization problems. We embed our technique into a genetic framework where our algorithm restarts with the set of new solutions formed by recombination and mutation of solutions found in the previous run. This method improves upon the existing Pareto local search algorithm for bi-objective and tri-objective quadratic assignment problem.In the second part of the thesis, we focus on non-preemptive scheduling algorithms. Here, we study the online problem of minimizing maximum stretch on a single machine. We present both positive and negative theoretical results. Then, we provide an optimally competitive semi-online algorithm. Furthermore, we study the problem of minimizing stretch on a single machine in a recently proposed rejection model. We show that there exists an O(1)-approximation ratio for minimizing average stretch. We also show that there exists an O(1)-approximation ratio for minimizing average flow time on a single machine. Lastly, we study the weighted average flow time minimization problem in online settings. We present a mathematical programming based framework that unifies multiple resource augmentation. Using the concept of duality, we show that there exists an O(1)-competitive algorithm for solving the weighted average flow time problem on unrelated machines. Furthermore, we proposed that this idea can be extended to minimizing l_k norms of weighted flow problem on unrelated machines.

The dermal filler Sculptra® has been employed to treat facial volume loss and age-related wrinkles in Europe since 1999. Sculptra® injections were administered 87,946 times (increase of 25.7% to 2012) in the USA in 2013. Except for histological analysis and clinical reports, data based on molecular biology or biochemistry, enlightening the mechanisms of action, do not exist to date. In contrast, such data are available for cross-linked hyaluronic acid, which is also administered for facial augmentation. To overcome this gap of knowledge, a comprehensive study about macroscopic to microscopic events occurring after Sculptra® injections was conducted. The augmentation of facial tissue with Sculptra® is approved; however, as the required repetitive biopsies could not be taken from the face, Sculptra® was injected to the inner side of the upper arms of 21 volunteers. Furthermore, this “off label use” was documented: The effect of the injections on the volunteer’s subjective quality of life was investigated using a questionnaire. For objective evaluation, the upper arms were photographed and sonographic measurements were applied. Photos of the treated upper arms revealed no changes in their shape during the study course. The amount of Sculptra® administered was not sufficient to augment the upper arms of postmenopausal women. Furthermore, these applied Sculptra® treatments of the upper arm did not improve the volunteers’ quality of life significantly. Upon sonographic measurement, however, a highly significant decrease in echogenicity was retrieved by comparing baseline subepidermal tissue values (t0) with 20 month (t2) values from either arm. Upon comparison of both treated sides, echogenicity was comparable; therefore 22 MHz sonography is an objective non-invasive measure to document the subcutaneous effect of Sculptra®. Immunofluorescence staining of sections from biopsies characterised the cell infiltrate and collagen type. CD68+ macrophages were found in direct proximity to PLLA, CD90+ fibroblasts aligned adjacently, while αSMA positive structures indicated myofibroblasts and neovascularisation. Substantial collagen type III deposition was detected right next to PLLA particles and collagen type I in the periphery of a given PLLA encapsulation. mRNA expression was strongly up-regulated for collagen type I and III transcripts, as well as for TGFß1 and TIMP1. PLLA particles were still retrievable 28 months after subcutaneous application. The augmenting effect of Sculptra® is generated by a complex reaction, comprised of various cells, chemokines and cytokines, leading to the proliferation of fibroblasts and their differentiation into myofibroblasts, synthesising a substantial amount of collagen in order to restore subcutaneous volume deficiencies. Degradation of facial and extra facial PLLA particles is considerably slower than described previously. The augmenting effect of Sculptra® diminishes over a period of 18-20 months in the face, but the degradation of PLLA particles seems to be much slower. Whether Sculptra® stimulates the synthesis of other ECM components, such as HA, or rationalises a continuous stimulus for collagen production, at least as long as it is not synthesised, should be analysed in further studies.