Catalogue Search | MBRL
Are you sure you want to remove the book from the shelf?
{{itemTitle}}
21
result(s) for
"Computer science Statistical methods Congresses."
Sort by:
Multivariate anomaly detection based on prediction intervals constructed using deep learning
It has been shown that deep learning models can under certain circumstances outperform traditional statistical methods at forecasting. Furthermore, various techniques have been developed for quantifying the forecast uncertainty (prediction intervals). In this paper, we utilize prediction intervals constructed with the aid of artificial neural networks to detect anomalies in the multivariate setting. Challenges with existing deep learning-based anomaly detection approaches include (
i
) large sets of parameters that may be computationally intensive to tune, (
ii
) returning too many false positives rendering the techniques impractical for use, and (
iii
) requiring labeled datasets for training which are often not prevalent in real life. Our approach overcomes these challenges. We benchmark our approach against the oft-preferred well-established statistical models. We focus on three deep learning architectures, namely cascaded neural networks, reservoir computing, and long short-term memory recurrent neural networks. Our finding is deep learning outperforms (or at the very least is competitive to) the latter.
Journal Article
Clustering-based adaptive data augmentation for class-imbalance in machine learning (CADA): additive manufacturing use case
Large amount of data are generated from in-situ monitoring of additive manufacturing (AM) processes which is later used in prediction modelling for defect classification to speed up quality inspection of products. A high volume of this process data is defect-free (majority class) and a lower volume of this data has defects (minority class) which result in the class-imbalance issue. Using imbalanced datasets, classifiers often provide sub-optimal classification results, i.e. better performance on the majority class than the minority class. However, it is important for process engineers that models classify defects more accurately than the class with no defects since this is crucial for quality inspection. Hence, we address the class-imbalance issue in manufacturing process data to support in-situ quality control of additive manufactured components. For this, we propose cluster-based adaptive data augmentation (CADA) for oversampling to address the class-imbalance problem. Quantitative experiments are conducted to evaluate the performance of the proposed method and to compare with other selected oversampling methods using AM datasets from an aerospace industry and a publicly available casting manufacturing dataset. The results show that CADA outperformed random oversampling and the SMOTE method and is similar to random data augmentation and cluster-based oversampling. Furthermore, the results of the statistical significance test show that there is a significant difference between the studied methods. As such, the CADA method can be considered as an alternative method for oversampling to improve the performance of models on the minority class.
Journal Article
Brainhack: a collaborative workshop for the open neuroscience community
Brainhack events offer a novel workshop format with participant-generated content that caters to the rapidly growing open neuroscience community. Including components from hackathons and unconferences, as well as parallel educational sessions, Brainhack fosters novel collaborations around the interests of its attendees. Here we provide an overview of its structure, past events, and example projects. Additionally, we outline current innovations such as regional events and post-conference publications. Through introducing Brainhack to the wider neuroscience community, we hope to provide a unique conference format that promotes the features of collaborative, open science.
Journal Article
Analysis and Optimal Control of Phase-Field Transition System
This e-book presents methods related to existence, uniqueness and regularity of solution, fractional steps, analysis of some boundary optimal control problems governed by phase-field transition system, conceptual algorithms to compute the approximate solution and boundary control. This volume should be a valuable reference for software engineers interested in modeling and simulation of phase transition processes.
eBook
The frailty model
Readers will find in the pages of this book a treatment of the statistical analysis of clustered survival data. Frailty models provide a powerful tool to analyze this data, and this book offers different methods based on these models.
eBook
Patient Participation at Health Care Conferences: Engaged Patients Increase Information Flow, Expand Propagation, and Deepen Engagement in the Conversation of Tweets Compared to Physicians or Researchers
Health care conferences present a unique opportunity to network, spark innovation, and disseminate novel information to a large audience, but the dissemination of information typically stays within very specific networks. Social network analysis can be adopted to understand the flow of information between virtual social communities and the role of patients within the network.
The purpose of this study is to examine the impact engaged patients bring to health care conference social media information flow and how they expand dissemination and distribution of tweets compared to other health care conference stakeholders such as physicians and researchers.
From January 2014 through December 2016, 7,644,549 tweets were analyzed from 1672 health care conferences with at least 1000 tweets who had registered in Symplur's Health Care Hashtag Project from 2014 to 2016. The tweet content was analyzed to create a list of the top 100 influencers by mention from each conference, who were then subsequently categorized by stakeholder group. Multivariate linear regression models were created using stepwise function building to identify factors explaining variability as predictor variables for the model in which conference tweets were taken as the dependent variable.
Inclusion of engaged patients in health care conference social media was low compared to that of physicians and has not significantly changed over the last 3 years. When engaged patient voices are included in health care conferences, they greatly increase information flow as measured by total tweet volume (beta=301.6) compared to physicians (beta=137.3, P<.001), expand propagation of information tweeted during a conference as measured by social media impressions created (beta=1,700,000) compared to physicians (beta=270,000, P<.001), and deepen engagement in the tweet conversation as measured by replies to their tweets (beta=24.4) compared to physicians (beta=5.5, P<.001). Social network analysis of hubs and authorities revealed that patients had statistically significant higher hub scores (mean 8.26×10-4, SD 2.96×10-4) compared to other stakeholder groups' Twitter accounts (mean 7.19×10-4, SD 3.81×10-4; t273.84=4.302, P<.001).
Although engaged patients are powerful accelerators of information flow, expanders of tweet propagation, and greatly deepen engagement in conversation of tweets on social media of health care conferences compared to physicians, they represent only 1.4% of the stakeholder mix of the top 100 influencers in the conversation. Health care conferences that fail to engage patients in their proceedings may risk limiting their engagement with the public, disseminating scientific information to a narrow community and slowing flow of information across social media channels.
Journal Article
Statistical Atlases and Computational Models of the Heart
This book constitutes the thoroughly refereed post-conference proceedings of the 5th International Workshop on Statistical Atlases and Computational Models of the Heart: Imaging and Modelling Challenges, STACOM 2014, held in conjunction with MICCAI 2014, in Boston, MA, USA, in September 2014. The 30 revised full papers were carefully reviewed and selected from numerous submissions. The papers cover a wide range of topics such as sections on cardiac image processing; atlas construction; statistical modelling of cardiac function across different patient populations; cardiac mapping; cardiac computational physiology; model customization; atlas based functional analysis; ontological schemata for data and results; integrated functional and structural analyses; as well as the pre-clinical and clinical applicability of these methods.
eBook
Finding hidden relevant documents buried in scientific documents by terminological paraphrases
Technical terms play an important role of effective queries for many users to search scientific databases. However, authors of scientific literature often employ alternative expressions to represent the meanings of specific terms, in other words,
Terminological Paraphrases
(TPs) in the literature for certain reasons, which leads to producing relevant documents that are not captured by conventional terms above. In this paper, we propose an effective way to retrieve “
de facto relevant documents
” which only contain those TPs and cannot be searched by conventional models in an environment with only controlled vocabularies by adapting Predicate Argument Tuple (PAT). The experiment confirms that PAT-based document retrieval is an effective and promising method to discover those kinds of documents and to improve the recall of terminology-based scientific information access models.
Journal Article
Training Students to Extract Value from Big Data
As the availability of high-throughput data-collection technologies, such as information-sensing mobile devices, remote sensing, internet log records, and wireless sensor networks has grown, science, engineering, and business have rapidly transitioned from striving to develop information from scant data to a situation in which the challenge is now that the amount of information exceeds a human's ability to examine, let alone absorb, it. Data sets are increasingly complex, and this potentially increases the problems associated with such concerns as missing information and other quality concerns, data heterogeneity, and differing data formats.
The nation's ability to make use of data depends heavily on the availability of a workforce that is properly trained and ready to tackle high-need areas. Training students to be capable in exploiting big data requires experience with statistical analysis, machine learning, and computational infrastructure that permits the real problems associated with massive data to be revealed and, ultimately, addressed. Analysis of big data requires cross-disciplinary skills, including the ability to make modeling decisions while balancing trade-offs between optimization and approximation, all while being attentive to useful metrics and system robustness. To develop those skills in students, it is important to identify whom to teach, that is, the educational background, experience, and characteristics of a prospective data-science student; what to teach, that is, the technical and practical content that should be taught to the student; and how to teach, that is, the structure and organization of a data-science program.
Training Students to Extract Value from Big Data summarizes a workshop convened in April 2014 by the National Research Council's Committee on Applied and Theoretical Statistics to explore how best to train students to use big data. The workshop explored the need for training and curricula and coursework that should be included. One impetus for the workshop was the current fragmented view of what is meant by analysis of big data, data analytics, or data science. New graduate programs are introduced regularly, and they have their own notions of what is meant by those terms and, most important, of what students need to know to be proficient in data-intensive work. This report provides a variety of perspectives about those elements and about their integration into courses and curricula.
eBook