Search

Deep learning (DL)’s dramatic rise in popularity across the domain sciences and industry has been accompanied by a correspondingly aggressive increase in the scale and computational complexity of DL workloads. In order to adopt state-of-the-art techniques, practitioners must wrestle with systems challenges of performance, cost, and scalability. In this dissertation, we identify the need for orchestration systems, which ease scaling burdens across the DL lifecycle through holistic, workload-aware optimizations. Drawing on both established techniques from data management research and new bespoke algorithms, we build practical orchestration engines to optimize three common DL workloads in the large-scale setting: model selection, data processing, and high-throughput serving. Our systems — which exploit workload- and context- specific opportunities — address a new layer of the large-scale DL optimization stack, more granular than current cluster managers and data systems, but still abstracted away low-level kernel & compiler optimizations. Empirical evaluations show that our orchestration techniques and systems can accelerate large-scale DL workloads by a large margin, even in complex, real-world settings. Our approach introduces a new technical lens, unifying systems, databases, and DL research, ultimately focused on democratizing and amplifying state-of-the-art DL innovations. Some of the systems proposed in this dissertation have already been adopted in production-scale industry pipelines, demonstrating the value of such orchestration optimizers for real-world DL.

Our work seeks to improve and adapt computing systems and machine learning (ML) algorithms to match each other’s requirements and capabilities. Due to the high computational demand for Deep Neural Networks (DNNs), accelerators such as GPUs are necessary to achieve low-latency inference. With GPUs getting more powerful, DNNs often fail to utilize a GPU’s parallelism fully. Understanding the GPU resources a DNN model can effectively use while still fulfilling the SLO of users allows us to run multiple DNN models concurrently by spatially sharing the GPU. Our DNN inference framework virtualizes the GPU, adapts to the DNN model, and improves the CPU-GPU coordination, to achieve a much higher aggregate inference throughput compared to other multiplexing techniques. While spatial sharing utilizes the GPU’s resources better, its static resource allocation is unsuitable for dynamic workloads. We propose a spatio-temporal scheduler that provides the right GPU resources for multiple DNNs and meets the inference tasks’ SLOs. Our spatio-temporal scheduler can run more models concurrently and achieve 4 × higher throughput than other scheduling techniques.Autotuning a DNN model customizes it to match the system’s capabilities. However, current autotuning frameworks do not consider a DNN model’s GPU demand. They produce a sub-optimally tuned model that does not provide the lowest possible latency when inferring with a smaller amount of GPU resources. Our enhanced autotuning produces a tuned model resilient to different amounts of GPU resources available at inference by targeting the appropriate GPU resources during tuning. Our framework enables concurrent autotuning of multiple models by using spatial multiplexing of the GPU, and eliminates several system overheads, thus decreasing tuning time by more than 70%.We apply our understanding of GPU to the task of localization in multi-user augmented reality (AR) applications. Our multi-user AR framework efficiently offloads AR computation to an edge server and lowers the localization time by 50%. Our implementation utilizes shared memory on the edge server to reduce the time for ’map merging’ between different users’ maps, a task that needs to be performed frequently with multi-user AR. Our approach cuts the map merging time by more than 80% compared to potential multi-user AR approaches.Our overall approach is to adapt computing resources to the algorithms that use them. It particularly benefits current ML algorithms and other applications that use these accelerators, such as AR. Our techniques can also improve the throughput of the newer generations of accelerators, which offer a significant speedup by using parallel compute engines.

N-Variant Execution (NVX) systems utilize artificial diversity techniques to enhance software security. The general idea is to run multiple different variants of the same program alongside each other while monitoring their run-time behavior. If a malicious input causes the execution paths of the diversified variants to diverge, the monitor can detect divergences, e.g. at the system call level, and take defensive action.Several NVX systems have been proposed over the last two decades, providing different security/performance characteristics. In general, security-oriented NVX systems greatly degrade performance, while high performance NVX systems have two disadvantages; they significantly increase the size of the Trusted Computing Base (TCB) or sacrifice security. In this dissertation, we want to investigate if it is possible to build an alternative NVX design that unifies the strengths of existing approaches.Security-oriented NVX systems are considered strong defenses that protect against a variety of attacks. However, a subset of modern attacks have the potential to bypass existing NVX systems. We identify limited available diversity as the main reason for that. Existing NVX systems execute diversified program variants on a single host. This means that the level of inter-variant diversity will be limited to what a single platform can offer.The main focus of this dissertation is to investigate the possibility of building the first distributed heterogeneous NVX system that executes program variants across multiple heterogeneous hosts. This approach can increase the level of internal diversity between the simultaneously running variants that can be supported, encompassing different instruction sets, endianness, calling conventions, system call interfaces, and differences in hardware security features. We expect that new challenges will arise from the distributed and heterogeneous nature of this design, however we believe that we will be able to provide sufficient solutions.

Legacy binaries need to continue functioning even when no source code has been preserved, to support the workflows of government and industry. The binaries often lack recent improvements in compiler design and software engineering practices, causing them to be slower and less secure than modern binaries. Binary rewriting seeks to patch, optimize, instrument, or harden binaries to bridge this gap, but existing practice is limited by the underlying static analysis. We created a framework, BinRec, to use dynamic analysis to lift binaries to LLVM IR then recompile them, which overcomes the limitations of static analysis.The protection of software against memory corruption exploits has a rich history, which this thesis both systematizes and extends. We present a study of the performance, precision, and security of control-flow integrity (CFI). Data-only attacks can bypass CFI, and so we present a defense against these attacks. The application of these hardening techniques to binaries deserves further study, and we discuss the extent to which security hardening can be applied to recompiled binaries.This dissertation presents building blocks for the securing of legacy binaries using dynamic analysis, which we hope will become a dominant paradigm in the secure software ecosystem of tomorrow.

Large distributed storage systems such as High Performance Computing (HPC) systems used by national or international laboratories must deliver performance for demanding scientific workloads that pose challenges from scale, complexity, and dynamism. Ideally, data is placed in locations and network routes are set to optimize performance, but the aforementioned challenges of large scientific systems and networks inhibits adaptive performance tuning on a large scale.If the data is not placed or routed with the potential future access patterns, the system risks having popular data stored on a slow file system. This can cause a steep drop in performance when access demand shifts to that data. Ideally, determining when and how to move data around in anticipation of future demand spikes can prevent steep drops in performance, and we refer to these drops as performance bottlenecks. We can leverage the fact that workloads shift through time, creating access patterns that can be used to forecast demand. Past accesses can be used to reveal important information when determining future accesses.An existing solution for determining where to move the data is to use access logs to create system performance models that can predict how the system reacts to fluctuations in demand. The drawback is that the search space of how to build an effective model is massive. The more there are locations where the data can be stored to larger the search space becomes. Additionally, applying the model to a system can potentially negatively impact any performance benefits of dynamically moving the data around the system. If a model is not set up correctly, the added overhead of using the model on the system may overshadow the benefit of applying the model on a target system. Ideally, the models created by the system engineers place and route the data to optimize performance, but workloads shift can be hard to predict using traditional methods such as heuristics.This dissertation applies and enhances online learning to predict and accelerate the performance of data movements and accesses in large scale storage systems. We focus on large scientific systems and networks like the CERN EOS, Pacific Northwest National Laboratory’s (PNNL) BlueSky system and the Caltech Tier 2 system. These systems allow for us to stress test our models with real data. Additionally it allows us to collect information about how accesses are distributed and executed on real systems. From the collected information we are able to create models based on real workloads. We are also able to reduce the training overhead of our models by determining which features may not bring added information to a model during training, and removing them. We developed a methodology to select features over time that is resilient to data collection noise added by new data collected from the system.Access logs created by system engineers often track hundreds of metrics describing its daily operation, and we must sift through these metrics to discover the most relevant metrics for the modeling task at hand. To explore different feature selection techniques, we have developed WinnowML, a system of automatically determining the most relevant feature subset for a specific modeling methodology when modeling system performance over time. We use WinnowML to determine what combination of existing techniques allow us to get a stable selected subset of features which will not vary significantly over time while keeping a low prediction error. Using the created list of features, system analysts can determine what features should be used when modeling an aspect of their system over time. Using WinnowML lowered the resulting mean absolute error by 13.6\% on average compared to the closest performing approach such as L1-regularization or Principal Component Analysis (PCA).To optimize the placement of data, we developed Geomancy, a tool that models the placement of data within a distributed storage system and reacts to drops in performance. Additionally to optimize the network routing decisions, we developed Diolkos, a tool that dynamically reroutes data flow in response to drops in performance. Using a combination of machine learning techniques suitable for temporal modeling, Geomancy and Diolkos determines when and where a bottleneck may happen due to changing workloads and suggests changes in the layout and routing decisions to mitigate or prevent them. Using WinnowML to determine which features to use when training our Geomancy tool, the predicted data layouts of Geomancy offered benefits for storage systems such as avoiding potential bottlenecks and increasing overall I/O throughput from 11\% to 30\%. It managed to free up resources that other workloads running concurrently could use. We then moved on to tackle the data transfer overhead in scientific networks. There exist several techniques that reroute data within scientific networks while improving network performance, leveraging latent parallel data transfer capabilities. The issue with these techniques is that they require a central authority, which may add computational and network overhead to the system. We propose a decentralized rerouting technique that exists at the switch level. When we applied Diolkos to the Caltech Tier 2 network, we found that our most accurate model, a dense model with one hidden layer trained using all the ports, increases throughput of switches up to 49\% compared to the best performing heuristic approach, exponentially weighted moving average, and up to 28\% when using a traditional controller.

In this poster we would like to present the new architecture and features of EVT 3 (Edition Visualization Technology), an open source Angular-based tool for the visualization of TEI-XML Digital Scholarly Editions. We will use and showcase two digital philology projects as test bed for the new functionalities.

Digital access to social sciences and humanities (SSH) research resources is necessary to enable broad use. For this purpose, a discovery platform for SSH resources was created. This paper describes its development, based on user, economic, and computer science research methods, and illustrates how such a platform supports social exchange.

This workshop uses the environment created around the HTR-United catalog to demonstrate and discuss how to build a dataset of ground truth for text recognition and document it, and how to use HTR-United and its suite of tools to control its quality and describe it in a standardized way.

The TEITOK REST API lets you interact with your corpus remotely - to upload documents in various formats - automatically converted to TEI/XML; to render search results; to run an NLP pipeline on the server; or download the content of corpus document, treat that content locally with local NLP tools or manual annotation tools, and then upload the results back to the server, where the new or corrected annotations will be incorporated into the original TEI/XML document without destroying any of the potentially complex annotations already in the document.

Engaging both creators of digital web-based content and researchers of content on the web, this tutorial introduces and demonstrates toolsets for crawling and sharing web archives. Unlike large-scale subscription services like ArchiveIt, these open-source tools allow individuals and small groups to archive websites and web-hosted data more granularly and deeply.

QTEI is based on two core components: web browser and JavaScript. The webbrowser, as one of the most comprehensive software packages for handling XML, is widely used and easy to maintain by end users. And JS is often part of the solution for rendering TEI/XML anyway, so with QTEI we created a framework that can fully implement a TEI web presence with these technologies.

We present the development of a richly annotated corpus of historical proceedings of the Carniolan Regional Assembly, which was the highest legislative body of the autonomy of the Carniola Herzogtum of the Habsburg Empire. We focus on the procedure used to convert the OCRed scans into Parla-CLARIN TEI encoded documents.

The poster presents the results of a workshop to be held during the annual DHd conference in March 2023. The workshop addresses the question how to strengthen the role of software and research software engineers in DH and how openness and collaboration can lead to more professionalization.

This paper explores the practices and experiences we have learned from developing an Open Science Infrastructure in the Asia Pacific. The paper purposes a unique three-pillar model; Advanced Technology to improve internet connectivity, a Knowledge Base on internet engineering, and Community Building across universities.

The improvement of the automatic transcription of manuscripts relies on an easier access to training data of good quality (ground truth). HTR-United offers a solution to find and document such datasets, potentially creating a common. We present the set-up of this ecosystem and its main outcomes.

Texas A&M University's Center of Digital Humanities Research (CoDHR) published the New Variorum Shakespeare (NVS) open-access in digital form, beginning with two editions, The Winter's Tale and A Midsummer Night's Dream in 2021, with plans to move forward with new editors who are under contract to create editions digitally in Corpora, and backward to eventually include all the volumes in the series, beginning with the first NVS Edition of Romeo and Juliet (1871). The CoDHR team has designed this project with three main goals: 1) to teach students and early career researchers the concepts behind variorum editing through interface design as well as tutorials; 2) to enable searching across and within volumes and variants using Modern English and major Act-Scene-Line numbers, returning not only exportable lists, text, and bibliographical citations, but also visualizations demonstrating everything from when characters speak in plays to the evolution of variant histories; and, in the final phase of the project, 3) to be interoperable with, and allow access to, other major Shakespeare digital resources including bibliographies of criticism, digital copies of editions published since Shakespeare's time, images, and videos. Following the practice of state-of-the-art digital humanities projects, we aim to render Shakespeare's texts and international criticism available world-wide.[1] To provide a little background, "The term variorum alludes to the Latin phrase editio cum notis variorum, that is, 'an edition with the notes of the various [editors and commentators],' a phrase indicating the chief purpose of a variorum edition: namely, to collect what has been written by various commentators, critics, and editors."[2] A Shakespeare variorum sets several editions of a play side by side in order to track sentence-, word-, and character-level changes. Each NVS volume comprises hundreds of pages, collates several editions, and takes editors decades--and very often a lifetime--to complete. These print editions also tend to be difficult to navigate for anyone who isn't a Shakespeare scholar or familiar with variorums. CoDHR's Digital NVS, however, organizes and presents variorums through an interface that is accessible, intuitive, and comprehensive. Unlike the print editions, this online application showcases all of the core components of a variorum at a glance: play text, commentary notes, and textual notes, making the content easier to understand. Additionally, new NVS editors will soon be able to use the web application's backend called Corpora to build editions, start to finish, using collation tools that will enable them to complete their work more quickly and accurately. What used to take editors years to complete can now be created through a single web application that allows for remote access and virtual collaboration. Having launched the site in Beta with the publication of The Winter's Tale and A Midsummer Night's Dream in July 2021, we are currently working on the next phases of the project as we move out of Beta: improving accessibility and usability; digitizing all NVS print editions from Romeo and Juliet (1871) to King Lear (2020); and assisting editors with creating born-digital NVS editions, using Corpora. We are also working on developing "a suite of research tools including an advanced search across multiple editions; visualizations of search results (such as a character-based frequency plot which will convey what characters most frequently use search terms in their speeches); the ability to sort and filter play lines, textual notes, and commentary in tabular format; and the ability to download raw versions of the data as JSON or TEI. We also hope to integrate external sources of information about a given play line or bibliographic entry."[3] It is our hope that the Digital NVS, which presents variorums in a new, more accessible way, will be an excellent resource not only for future NVS editors, but for readers, scholars, educators, and performers. Bibliography Burdick, Anne and Tarpley, Bryan. "NVS Tools." New Variorum Shakespeare, Beta Release. July 2021. https://newvariorumshakespeare.org/tools/. Accessed Dec. 5, 2021 Burdick, Anne, Torabi, Katayoun, Bryan Tarpley, and Mandell, Laura. "Using Data and Design to Bring the New Variorum Shakespeare Online." The Routledge Handbook of Shakespeare and Interface. Edited by Paul Budra and Clifford Werier, 2022. Knowles, Richard. Shakespeare Variorum Handbook: A Manual of Editorial Practice. 2nd ed. Committee on the New Variorum Edition of Shakespeare of the Modern Language Association of America, 2003. Torabi, Katayoun. "How to Digital NVS." New Variorum Shakespeare, Beta Release. July 2021 https://newvariorumshakespeare.org/how-to/.nd. Accessed Dec. 5, 2021 Werstine, Paul. "NVS History." New Variorum Shakespeare, Beta Release. July 2021. https://newvariorumshakespeare.org/about/. Accessed Dec. 5, 2021 [1] Information about the Digital NVS (overview of the project, goals, uses, etc.) can be found in "Using Data and Design to Bring the New Variorum Shakespeare Online" in The Routledge Handbook of Shakespeare and Interface, 2022. [2] Information about the meaning of variorum and the history of the NVS can be found in Paul Werstine's "NVS History," New Variorum Shakespeare, Beta Release. July 2021. [3] Anne Burdick and Bryan Tarpley, "NVS Tools," New Variorum Shakespeare, Beta Release. July 2021; and Burdick, et al., "Using Data and Design to Bring the New Variorum Shakespeare Online" in The Routledge Handbook of Shakespeare and Interface. Eds. Paul Budra and Clifford Werier, 2022.

The contribution highlights the importance of developing DH methods for large-scale investigation of depicted materials and surface qualities in medieval painting. Preliminary findings from our project, which is developing methods for AI-based recognition of materials in digitized paintings and evaluating their application to address art historical research questions, are presented.

The poster will introduce and share the first results of the Horizon Europe project MEMORISE, which creates a framework to preserve and enhance Heritage related to Nazi Persecution by virtualizing and linking multimodal data resources and by developing and offering novel digital technologies for accessing, exploring and engaging with HNP.

We propose to leverage public vulnerability scanners to examine the servers where online digital projects are hosted and list their common vulnerabilities. Provided in an anonymized, aggregated fashion, this assessment will shed light on our hypothesis: cyber threats are a significant factor in the degradation of online projects.

In this panel we offer a multifaceted and interdisciplinary perspective on the development of the Sloane Lab. We contextualise it both within a wider, nationally-funded drive to make the UK's cultural heritage, and information about it, computationally tractable and the more internationally-positioned work of 'Collections as Data'.

Manu McFrench is a handwritten text recognition (HTR) model dedicated to French texts from the modern and contemporary periods. It was trained using data generated through the CREMMA infrastructure and/or shared via the HTR-United catalog. We demonstrate its usefulness to reduce the costs of training such models for specific handwritings.

This paper presents a platform providing a digital census of copies of a limited number of printed editions and mapping their owners and readers' annotations represented in an RDF framework. Its ResearchSpace instance, magnetic-margins.com, provides interactive statistical analyses of the data that is encoded predominantly using CIDOC-CRM and FRBRoo.

Participants are invited to explore modelling and annotation through exercises demonstrating music research conducted in Oxford and Vienna. After hands-on ontology design exercises with pen and paper, they are introduced to cutting-edge digital tooling and led through research processes of an ongoing project investigating the Vienna Philharmonic's New Year's Concerts.

This workshop will discuss the field of stylometry and authorship attribution, and its application to the development of legal evidence. Workshop participants will be invited to participate in a discussion of the development of consensus standards to help improve reliability and to permit evaluation of proposed stylometric evidence.

In this poster we present the ongoing development of an integrated free and open source toolchain for semantic annotation of digitised cultural heritage. The toolchain development involves the specification of a common data model that aims to increase interoperability across diverse datasets and to enable new collaborative research approaches.

As technology scales, transistors become smaller and aggressive power optimization techniques combined with high operation frequencies and performance-enhancing microarchitectural techniques are employed to achieve increasingly higher performance and power efficiencies. Unfortunately, these developments make the modern systems more vulnerable to soft errors, which are becoming a critical issues in both hardware and software domains. Motivated by this observation, in this work, we propose, implement, and evaluate two error propagation metrics in order to characterize error propagation at both software and hardware levels. The first metric aims to measure error propagation on program data structures, whereas the second one measures the fraction of corrupted locations in the cache memory structure for a given period of time. We evaluate our proposed metrics by performing an empirical study of two application programs using both single-threaded and multi-threaded executions, and varying various experimental parameters such as thread count, error rate, location of errors, and architectural parameters. Our extensive experimental analysis reveals that error propagation over program data structures is highly dependent on application behavior.Further, depending on the cache parameters used, propagation of errors on cache can exhibit different patterns. This paper also discusses how our observed error propagation trends in program data structures and data caches are correlated with each other, focusing in particular on the differences in error propagation speeds in application data structures and data caches.

Addressing compromised device is a problem for virtually all largeorganizations. Compromised devices can propagate malware resulting intheft of computing resources, loss of sensitive data, and extortion ofmoney. Unfortunately, large organizations do not have an oracle intodevice compromise. Instead, organizations must address compromisewithout straightforward answers to critical questions such as: "Isthis device compromised?", "Why/How is this device compromised?","What's the correct intervention?". This problem, in part, resultsfrom limited observational vantage points, differences in interventioncapabilities, and evolving adversaries with differing incentives. Inthis dissertation, I develop systems that empirically address multipletypes of device compromise using large-scale observations withindifferent organizations, thus placing us on a stronger footing todevise better interventions.I first describe an approach used at Facebook for detecting maliciousbrowsers extensions. I present a methodology whereby users exhibitingsuspicious online behaviors are scanned (with permission) to identifyextensions in their browsers, and those extensions are in turn labeledbased on the threat indicators they contain. Employing thismethodology at Facebook I identify more than 1,700 lexically distinctmalicious extensions, and use this labeling to drive user deviceclean-up efforts as well notify browser vendors.Next, I examine for-profit services offering to artificiallymanipulate a user's social standing on Instagram. I identify thetechniques used by these services to drive social actions, detail howthey are structured to evade straightforward detection, andcharacterize the dynamics of their customer base. Finally, I constructcontrolled experiments to disrupt these services and analyze howdifferent approaches to intervention can drive different reactions,thus providing distinct trade-offs for defenders.Lastly, I describe a large-scale measurement of 15,000 laptop anddesktop devices on a university's network to characterize theprevalence of security "best practices" and security-relevantbehaviors, and quantify how they relate to device compromise. I usepassive network traffic analysis techniques to infer a broad range ofdevice features and per-machine compromise state. I find a number ofbehaviors positively correlate with host compromise, and few "bestpractices" exhibit negative correlations that would support theirvalue in improving end user security.

Reproducibility is the cornerstone of the scientific method. Yet, in the computational and data-intensive branches of science, a gap exists between current practices and the ideal of having every new scientific discovery be _easily_ reproducible. At the root of this problem are the dysfunctional forms of communication between the distinct stakeholders of science: researchers, their peers, students, librarians and other consumers of research outcomes working in ad-hoc ways; groups of individuals organized as independent silos, sharing minimal information between them, all of them with the common task of publishing, obtaining, re-executing and validating experimentation pipelines associated to scientific claims contained in scholarly articles and technical reports. This dissertation characterizes the practical challenges associated to the research lifecycle (creation, dissemination, validation, curation and re-use of scientific explorations) and draws analogies with similar problems experimented by software engineering communities in the early 2000's. DevOps, the state-of-the-art software delivery methodology followed by companies and open source communities, appeared in late 2000's and addresses these analogous issues. By framing the problem of research delivery (iterating the research lifecycle) as a problem of software delivery, it becomes possible to repurpose the DevOps methodology to address the practical challenges faced by experimenters across the domains of computational and data-intensive science. This thesis presents Popper, an experimentation protocol for writing articles and carrying out scientific explorations following DevOps principles. Popper brings agility to research delivery in a domain-agnostic way by considering the end-to-end research delivery cycle, making it easier for all the stakeholders of science to publish, access, re-execute and validate experimentation pipelines. This dissertation also presents reusable tools in the domain of computer systems research. In a research delivery context, this toolset covers the multiple phases of the research lifecycle and helps systems research practitioners carry out validations of scientific claims in this domain in an agile manner. For this work, I was awarded the 2018 Better Scientific Software Fellowship, an initiative from the US Department of Energy whose goal is to foster best software development practices among the US scientific community. The Popper CLI tool, and its associated education materials, are currently being used to train new generations of scientists on how to be aware and guard against the multiple practical challenges in reproducibility.

Cancer is a fatal disease and one of the leading causes of death worldwide. The cure rate in cancer treatment remains low; hence, cancer treatment is gradually shifting toward personalized treatment. Artificial intelligence (AI) and radiomics have been recognized as one of the potential areas of research in personalized medicine in oncology. Several researchers have identified the capabilities of AI and radiomics to characterize phenotype and there by predict the outcome of treatment in oncology. Although AI and radiomics have shown promising initial results in diagnosis and treatment in oncology, these technologies are also facing challenges of standardization and scalability. In the last few years, researchers have been trying to develop a research infrastructure for federated machine learning that increases the usability of Big Data for clinical research. These research infrastructures are based on the findable, accessible, interoperable, and reusable (i.e., FAIR) data principles. The India-Dutch "big imaging data approach for oncology in a Netherlands India collaboration" (BIONIC) is a jointly funded initiative by the Dutch Research Council (NWO) and the Indian Ministry of Electronics and Information Technology (MeitY), aiming to introduce radiomic-based research into clinical environments using federated machine learning on geographically dispersed collections of FAIR data. This article described a prototype end-to-end research infrastructure implemented through the BIONIC partnership into a leading cancer care public hospital in India.

The data within collections from all Digital Humanities fields must be trustworthy. To this end, both provenance and change-tracking systems are needed. This contribution offers a systematic review of the metadata representation models for provenance in RDF, focusing on the problem of modelling conjectures in humanistic data., Comment: 10 pages, 2 figures, submitted to the ADHO Digital Humanities Conference 2023

Many complex applications require the solution of initial-value problems where some components change fast, while others vary slowly. Multirate schemes apply different step sizes to resolve different components of the system, according to their dynamics, in order to achieve increased computational efficiency. The stiff components of the system, fast or slow, are best discretized with implicit base methods in order to ensure numerical stability. To this end, linearly implicit methods are particularly attractive as they solve only linear systems of equations at each step. This paper develops the Multirate GARK-ROS/ROW (MR-GARK-ROS/ROW) framework for linearly-implicit multirate time integration. The order conditions theory considers both exact and approximative Jacobians. The effectiveness of implicit multirate methods depends on the coupling between the slow and fast computations; an array of efficient coupling strategies and the resulting numerical schemes are analyzed. Multirate infinitesimal step linearly-implicit methods, that allow arbitrarily small micro-steps and offer extreme computational flexibility, are constructed. The new unifying framework includes existing multirate Rosenbrock(-W) methods as particular cases, and opens the possibility to develop new classes of highly effective linearly implicit multirate integrators.

Mathematical model-based analysis and control of human immunodeficiency virus (HIV) infection have recently provided important advantages in medicine. In this paper, firstly the literature on mathematical models and applied control methods will be surveyed to evaluate the HIV models and therapy. Secondly, a cubature Kalman filter-based nonlinear model predictive control is proposed for the multi-input multi-output control of HIV infection for decreasing the cost of sensory devices and increasing the efficiency of therapy. By doing so both unmeasurable states and personalized parameters of the HIV infection are jointly estimated in a control process to generate suitable drug dosages. In the literature, the applied drug dosages are in continuous or on/off levels. For a practical application of continuous drug dosage-level, it has been discretized into 10 levels of full dosage level. Therefore, the applied drug dosages are in piecewise-continuous levels instead of continuous values or on/off levels. The proposed observer-controller configuration has been applied to the strong and moderate therapy levels of long-term non-progressive as well as fast-progressive patients with personalized parameters, where the application results are discussed for 1-, 5-, 10- and 20-year periods. The computational results show that satisfactory performances are obtained for future applications in terms of the root-mean-squared error of the estimation and control, and in terms of the integral sum of the control input.

ispartof: ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE vol:116 status: published

Line planning is one of the important steps in the public transit planning process. It is a difficult combinatorial problem with an enormous search space. For networks from practice, which are typically large and have more complex constraints, it takes an unreasonable amount of time to find good solutions with the current methods. When the objective is to minimize passenger travel time, most of the calculation time is spent on solving the passenger routing sub-problem. This paper proposes a local evaluation technique to significantly speed up this most critical component. A change is assumed to have its biggest impact on passengers already travelling close to the change. This idea is implemented by putting a grid over the network and only re-evaluating the part of the network around the change, while still taking into account all passengers. This results in a smaller computation time for each passenger routing, with only a limited loss of quality, and it allows more iterations. On a number of benchmark instances, the results obtained by the grid approach are compared to the same algorithm without any local evaluation. The grid approach is significantly faster and allows to perform 5–20 times more evaluations per second, while having a minimal loss of quality per evaluation. When the available calculation time is limited, the grid based approach obtains the best solutions. On the larger benchmark instances, the grid based approach can use more complex neighborhoods resulting in even better solutions. The method is very flexible and can be integrated in many line planning algorithms and probably even in algorithms for other network problems.

Background Metabolomics has been proven to be an invaluable research tool by providing comprehensive insight into systemic metabolism. However, the lack of scalable and quantitative methods with known reference intervals (RIs) and documented reproducibility has prevented the use of metabolomics in the clinical setting. Objective The objective of this study was to validate the developed quantitative nuclear magnetic resonance (NMR) spectroscopy-based metabolomics platform for canine serum and plasma samples and determine optimal sample handling conditions for its use. Methods Altogether, 8247 canine samples were analyzed using a Bruker's 500 MHz NMR spectrometer. Using statistical approaches derived from international guidelines, we studied method precision, measurand stability in various long- and short-term storage conditions, as well as the effect of prolonged contact with red blood cells (RBCs), and differences among blood collection tubes. We also screened interferences with lipemia, hemolysis, and bilirubinemia. The results were compared against routine clinical chemistry methods, and RIs were defined for all measurands. Results We determined RIs for 123 measurands, most of which were previously unpublished. The reproducibility of the results of the NMR platform appeared generally outstanding, and the integrity of the results can be ensured by following standard blood drawing and processing guidelines. Conclusions Owing to the advantages of quantitative results, high reproducibility, and scalability, this canine metabolomics platform holds great potential for numerous clinical and research applications to improve canine health and well-being.

A novel coding strategy is proposed for a broadcast setting with two transmitter (TX) antennas and two single-antenna receivers (RX). The strategy consists of using space-time block coding to send a common message (to be decoded by both RXs) across the two TX antennas, while each TX antenna also sends a private message to one of the RXs. The relative weight of the private and common messages from each TX antenna is tuned to maximize the instantaneous achievable sum-rate of the channel. Closed-form expressions for the optimal weight factors are derived. In terms of the generalized degrees of freedom (GDoF) metric, the new scheme is able to achieve the sum-GDoF with finite precision channel state information at the transmitter (CSIT) of the two user broadcast channel. Moreover, as opposed to the existing rate-splitting schemes, the proposed scheme yields instantaneous achievable rates that are independent of the channel phases. This property is instrumental for link adaptation when only magnitude CSIT is available. Our numerical results indeed demonstrate the superiority of the scheme for the 2-user setting in case of magnitude CSIT. Extension to a more generalK-user scenario is briefly discussed.

Objective Coronary artery calcium (CAC) score is a strong predictor for future adverse cardiovascular events. Anthropomorphic phantoms are often used for CAC studies on computed tomography (CT) to allow for evaluation or variation of scanning or reconstruction parameters within or across scanners against a reference standard. This often results in large number of datasets. Manual assessment of these large datasets is time consuming and cumbersome. Therefore, this study aimed to develop and validate a fully automated, open-source quantification method (FQM) for coronary calcium in a standardized phantom.Materials and Methods A standard, commercially available anthropomorphic thorax phantom was used with an insert containing nine calcifications with different sizes and densities. To simulate two different patient sizes, an extension ring was used. Image data were acquired with four state-of-the-art CT systems using routine CAC scoring acquisition protocols. For interscan variability, each acquisition was repeated five times with small translations and/or rotations. Vendor-specific CAC scores (Agatston, volume, and mass) were calculated as reference scores using vendor-specific software. Both the international standard CAC quantification methods as well as vendor-specific adjustments were implemented in FQM. Reference and FQM scores were compared using Bland-Altman analysis, intraclass correlation coefficients, risk reclassifications, and Cohen's kappa. Also, robustness of FQM was assessed using varied acquisitions and reconstruction settings and validation on a dynamic phantom. Further, image quality metrics were implemented: noise power spectrum, task transfer function, and contrast- and signal-to-noise ratio among others. Results were validated using imQuest software.Results Three parameters in CAC scoring methods varied among the different vendor-specific software packages: the Hounsfield unit (HU) threshold, the minimum area used to designate a group of voxels as calcium, and the usage of isotropic voxels for the volume score. The FQM was in high agreement with vendor-specific scores and ICC's (median [95% CI]) were excellent (1.000 [0.999-1.000] to 1.000 [1.000-1.000]). An excellent interplatform reliability of kappa = 0.969 and kappa = 0.973 was found. TTF results gave a maximum deviation of 3.8% and NPS results were comparable to imQuest.Conclusions We developed a fully automated, open-source, robust method to quantify CAC on CT scans in a commercially available phantom. Also, the automated algorithm contains image quality assessment for fast comparison of differences in acquisition and reconstruction parameters.

Drug delivery devices are promising tools in the pharmaceutical field, as they are able to maximize the therapeutic effects of the delivered drug while minimizing the undesired side effects. In the past years, electrospun nanofibers attracted rising attention due to their unique features, like biocompatibility and broad flexibility. Incorporation of active principles in nanofibrous meshes proved to be an efficient method for in situ delivery of a wide range of drugs, expanding the possibility and applicability of those devices. In this review, the principle of electrospinning and different fields of applications are treated to give an overview of the recent literature, underlining the easy tuning and endless combination of this technique, that in the future could be the new frontier of personalized medicine.

In frequency domain Electromagnetic Induction (EMI) surveys, an image of the electrical conductivity of the subsurface is obtained non-invasively. The electrical conductivity can be related to important subsurface properties such as the porosity, saturation or water conductivity via Archie’s law. The advantage of geophysical EMI surveys is its cost-effectiveness because it is a non-contacting method, one can easily walk with the device or mount in on a vehicle or a helicopter (AEM).The process of finding the conductivity profile from the collected field data is an ill-posed inverse problem. Regularization improves the stability of the inversion and, based on Occam’s razor principle, a smoothing constraint is typically used with a very large number of thin layers. However, the conductivity profiles are not always expected to be smooth. Another alternative is to use a predefined number of layers and to invert for their conductivity and thickness. This can yield sharp contrasts in conductivity. In practice however, the real underground might be either blocky or smooth, or somewhere in between. Those standard constraints are thus not always appropriate.We develop a new minimum-structure inversion scheme in which we transform the model into the wavelet space and impose a sparsity constraint. This sparsity constrained inversion scheme minimizes an objective function with a least-squares data misfit and a sparsity measure of the model in the wavelet domain. With a solid understanding of wavelet theory, a novel and intuitive model misfit term was developed, allowing for both smooth and blocky models, depending on the chosen wavelet basis. A model in the wavelet domain has both temporal (i.e. low and high frequencies) and spatial resolution, and penalizing small-scale coefficients effectively reduces the complexity of the model.Comparing the novel scale-dependent wavelet-based regularization scheme with wavelet-based regularization with no scale-dependence, revealed significantly better results (Figure A and B) w.r.t. the true model. Comparing with standard Tikhonov regularization (Figure C and D) shows that our scheme can recover high amplitude anomalies in combination with globally smooth profiles. Furthermore, the adaptive nature of the inversion method (due to the choice of wavelet) allows for high flexibility because the shape of the wavelet can be exploited to generate multiple representations (smooth, blocky or intermediate) of the inverse model.We have introduced an alternative inversion scheme for EMI surveys that can be extended to any other 1D geophysical method. It involves a new model misfit or regularization term based on the wavelet transform and scale-dependent weighting which can easily be combined with the existing framework of deterministic inversion (gradient-based optimization methods, L-curve criterion for optimal regularization parameter). A challenge remains to select the optimal wavelet, however, the ensemble of inversion results with different wavelets can also be used to qualitatively assess uncertainty.

Sampling from the lattice Gaussian distribution has emerged as a key problem in coding and decoding while Markov chain Monte Carlo (MCMC) methods from statistics offer an effective way to solve it. In this paper, the sliced lattice Gaussian sampling algorithm is proposed to further improve the convergence performance of the Markov chain targeting at lattice Gaussian sampling. We demonstrate that the Markov chain arising from it is uniformly ergodic, namely, it converges exponentially fast to the stationary distribution. Meanwhile, the convergence rate of the underlying Markov chain is also investigated, and we show the proposed sliced sampling algorithm entails a better convergence performance than the independent Metropolis-Hastings-Klein (IMHK) sampling algorithm. On the other hand, the decoding performance based on the proposed sampling algorithm is analyzed, where the optimization with respect to the standard deviation $\sigma >0$ of the target lattice Gaussian distribution is given. After that, a judicious mechanism based on distance judgement and dynamic updating for choosing $\sigma $ is proposed for a better decoding performance. Finally, simulation results based on multiple-input multiple-output (MIMO) detection are presented to confirm the performance gain by the convergence enhancement and the parameter optimization.

Thousands of new users join social media website everyday, generating huge amounts of new data. Twitter users for example, generate millions of new posts per day. This can flood our users with huge amounts of information, and thus overload them with information that for the most part they are not interested in. To fix this problem, we need to only show our users information relative to them, such as posts from people they are following. This thesis focuses on how to make accurate recommendations to each user, on which users/pages to follow, thus helping the user view information that is important to them. In particular, we focus on exploring the following research questions: 1) which features yield the best recommendation accuracy, and 2) given those features, what is the best granularity for them, that captures the underlying dynamics, leading to high accuracy.

The current Internet architecture requires applications to transparently bind and manage network addresses and ports. This design creates and exacerbates several problems for the current Internet as well as its future evolution. These problems and challenges include (1) network address mobility and multihoming; (2) future Internet evolution; (3) service- or content-centricity; and (4) adding, subtracting, or evolving the layers in the network stack.To address all of these problems, this thesis introduces a novel approach to Internet naming and addressing, which we call hidden identifiers. Hidden identifiers enable applications to semantically express the exact network resource they desire, and allows the operating system to subsequently bind and manage all other network concerns. In this manner, we provide integrated support for all the problem-cases enumerated above while simultaneously simplifying the network API presented to applications.We introduce, implement, and evaluate HIDRA, the first network stack architecture based on hidden identifiers. We show that this network stack provides integrated support for features such as network mobility and multihoming, and explain how the HIDRA architecture can support a wide range of semantic bindings. Finally, we show how this feature can be leveraged to support the goals of information-centric networking on top of the existing TCP/IP network stack by slightly modifying DNS and HTTP.

Universities in Indonesia already have academic information systems. This university in developing its academic information system is carried out individually according to conditions and needs. Currently there are no guidelines in providing direction for the planning, development, implementation and service of academic information systems based on information and communication technology in universities. In addition to this, it is based on the planning and development of an academic information system which is what will be planned and developed and how its development guidelines are not yet available. The purpose of this study is to provide guidance in building academic information systems at private universities that can be applied in South Sumatra Province. This goal needs to be done because in general the academic information system business processes and services as well as the need for information in each university have in common namely the acceptance of new students, academic management and academic release. The result of the research is an academic information system architecture as a guideline in planning the development of an academic information system according to the conditions of private universities in South Sumatra Province. This study uses the EAP method with the steps of activities including planning initiation, business modeling and survey of current institutions, systems and technologies, data architecture and application architecture. The results of this study are a guide in building a system in the form of an academic information system architecture that can be applied to private universities in Indonesia especially South Sumatra Province.This study uses the EAP / Enterprise Architecture Planning method with activity steps including planning initiation, business modeling and institutional surveys, current systems and technology, data architecture and application architecture.

Graphical abstract, The interactions between cells and their extracellular matrix (ECM) are critically important for homeostatic control of cell growth, proliferation, differentiation and apoptosis. Transmembrane integrin molecules facilitate the communication between ECM and the cell. Since the characterization of integrins in the late 1980s, there has been great advancement in understanding the function of integrins at different subcellular levels. However, the versatility in molecular pathways integrins are involved in, the high diversity in their interaction partners both outside and inside the cell as well as on the cell membrane and the short lifetime of events happening at the cell–ECM interface make it difficult to elucidate all the details regarding integrin function experimentally. To overcome the experimental challenges and advance the understanding of integrin biology, computational modeling tools have been used extensively. In this review, we summarize the computational models of integrin signaling while we explain the function of integrins at three main subcellular levels (outside the cell, cell membrane, cytosol). We also discuss how these computational modeling efforts can be helpful in other disciplines such as biomaterial design. As such, this review is a didactic modeling summary for biomaterial researchers interested in complementing their experimental work with computational tools or for seasoned computational scientists that would like to advance current in silico integrin models.

The main objective of the present paper is to report two studies on mathematical and computational techniques used to model the behaviour of the aorta in the human cardiovascular system. In this paper, an account of the design and implementation of two distinct models is pre-sented: a Windkessel model and an agent-based model. Windkessel model represents the left heart and arterial system of the cardiovascular system in the physiological domain. The agent-based model offers a simplified account of arterial behaviour by randomly generating arterial parameter values. This study has described the mechanism how and when the left heart contracts and pumps the blood out of the aorta, and it has taken the Windkessel model one step further. The results of this study show that the dynamics of the aorta can be explored in each modelling approaches as proposed and implemented by our research group. It is thought that this study will contribute to the literature in terms of development of the Windkessel model by considering its timing and redesigning it with digital electronics perspective.

Cells regulate themselves via dizzyingly complex biochemical processes called signaling pathways. These are usually depicted as a network, where nodes represent proteins and edges indicate their influence on each other. In order to understand diseases and therapies at the cellular level, it is crucial to have an accurate understanding of the signaling pathways at work. Since signaling pathways can be modified by disease, the ability to infer signaling pathways from condition- or patient-specific data is highly valuable. A variety of techniques exist for inferring signaling pathways. We build on past works that formulate signaling pathway inference as a Dynamic Bayesian Network structure estimation problem on phosphoproteomic time course data. We take a Bayesian approach, using Markov Chain Monte Carlo to estimate a posterior distribution over possible Dynamic Bayesian Network structures. Our primary contributions are (i) a novel proposal distribution that efficiently samples sparse graphs and (ii) the relaxation of common restrictive modeling assumptions. We implement our method, named Sparse Signaling Pathway Sampling, in Julia using the Gen probabilistic programming language. Probabilistic programming is a powerful methodology for building statistical models. The resulting code is modular, extensible, and legible. The Gen language, in particular, allows us to customize our inference procedure for biological graphs and ensure efficient sampling. We evaluate our algorithm on simulated data and the HPN-DREAM pathway reconstruction challenge, comparing our performance against a variety of baseline methods. Our results demonstrate the vast potential for probabilistic programming, and Gen specifically, for biological network inference. Find the full codebase at https://github.com/gitter-lab/ssps, Comment: 15 pages, 10 figures; added Appendices C and D