1,720,967 research outputs found
Deep Learning Under Constraints: Techniques for Data-Limited Medical and Recommender System Applications
Get PDFData is the fundamental resource required for training deep learning models. The availability and collection of massive datasets, combined with increasing computational power, has enabled the training of large models that learn complex tasks with human-level proficiency. However, the abundance of data is not universal across all domains. This thesis addresses data scarcity issues in deep learning, focusing on two domains: medicine and recommender systems. In medicine, data scarcity stems from the prohibitive costs of annotating medical data for model training and stringent privacy regulations that limit data sharing. Conversely, recommender systems face cold-start problems, which arise when a new user or item enters the system without sufficient historical data for informed recommendations. To reduce the dependency on costly electrocardiogram (ECG) annotations for training heart arrhythmia classifiers, we employed transfer learning methods to teach the models foundational knowledge about ECG signals. We developed novel pre-training tasks to acquire this knowledge using one of the largest public ECG databases. By fine-tuning these pre-trained models on small datasets, we improved performance across specialized ECG classification tasks by up to +6.57% over models trained from scratch. To address privacy concerns regarding data sharing, we employed federated learning, which enables collaborative model training among medical institutions by sharing locally computed updates to the model instead of raw training data. In a large study using a diverse collection of ECG databases, we demonstrated that ECG classifiers trained via federated learning outperform models trained in isolation on local data. Furthermore, on out-of-distribution data from foreign institutions, these classifiers nearly match the performance of models trained with shared data. Addressing item cold-start issues in sequential recommender systems, we developed a novel dynamic storage model that maintains a latent memory of user-item interactions without requiring gradient updates or side information, ensuring efficient representation and timely recommendation of new items during inference. Our proposed model outperforms a similar approach by more than 29% on recommendation tasks and demonstrates the ability to adapt to entirely new datasets without any fine-tuning in zero-shot transfer scenarios. Collectively, our contributions tackle various data scarcity issues in selected domains, advancing contemporary deep learning methodologies in data-limited scenarios
Interpretierbare Deep Learning-Ansätze für die Erkennung von Biomarkern aus hochdimensionalen biomedizinischen Daten
No full textWith the advances in high-throughput data acquisition technologies, the amount of heterogeneous and complex data is constantly increasing. The application of intelligent algorithms such as deep neural networks (DNNs), which learn a hierarchy of increasingly complex features from the data, is emerging as an effective paradigm for analyzing complex datasets. In medical research, however, deep learning (DL) may suffer from overfitting due to the high dimensionality of the data. The scarcity of good quality labeled data also intensifies this issue due to the expensive and time-consuming process of providing labels and metadata by the human expert.
Besides, despite the simple linear operation of core building blocks of DNNs, the hierarchical combination of these blocks may result in over-parameterization, which makes it challenging to explain their behavior. The black-box nature of these models raises a severe issue on the trustworthiness and reliability of deployed models, especially in high-stakes prediction applications. Therefore, to analyze high-dimensional medical data using DL models in medical settings, we need to address two important questions: 1) How to deal with the curse of dimensionality and limitation of annotated data? 2) How to improve the transparency of deep learning models through interpretability, as it potentially leads to a better understanding of the data and the deployed model?
This thesis addresses these challenges in high-throughput structured data and high-dimensional imaging data modalities. We begin our study on high-throughput structured data with the application of proteomics data analysis. We robustly learn the data representation and extract the medically relevant information using DL techniques. We develop novel data analysis based on what the DL model can learn through interpreting its predictions. This information enables getting insight into the data patterns and discovering discriminating features. We also justify the reliability of the model interpretation through comprehensive quantitative assessments. We show that the proper combination of DL techniques coupled with interpretation strategies that enable an in-depth understanding of model decisions can guide towards a reliable clinical decision support system.
Further, we study DL techniques on high-dimensional imaging data. Unlike structured data where desired features appear with slight deviations, regions of interest on medical images may appear with a large deviation on different data points. Therefore, we built our image analysis on supervised convolutional neural networks (CNN), which can handle large deviations.
We investigate different CNN architectures and compare their strength. Finally, we built a robust pipeline on heterogeneous imaging data with the challenging application of human spinal vertebra detection-identification.
To deal with the scarcity of data, we show how different techniques, including transfer learning, data augmentation, human-in-the-loop, and synthetic generation of data in medical settings, boost generalization.Mit den Fortschritten bei den Technologien zur Datenerfassung mit hohem Durchsatz nimmt die Menge heterogener und komplexer Daten ständig zu. Die Anwendung intelligenter Algorithmen wie deep neural networks (DNNs), die eine Hierarchie zunehmend komplexer Merkmale aus den Daten lernen, entwickelt sich zu einem effektiven Paradigma für die Analyse komplexer Datensätze. In der medizinischen Forschung kann deep learning (DL) jedoch aufgrund der hohen Dimensionalität der Daten unter einer Überanpassung leiden. Der Mangel an qualitativ hochwertigen beschrifteten Daten verschärft dieses Problem noch, da die Bereitstellung von Beschriftungen und Metadaten durch einen menschlichen Experten teuer und zeitaufwändig ist. Außerdem führt die hierarchische Kombination dieser Blöcke trotz der einfachen linearen Funktionsweise der Kernbausteine von DNNs zu einer Überparametrisierung, die es schwierig macht, ihr Verhalten zu erklären. Um hochdimensionale medizinische Daten mit fortschrittlichen DL-Modellen in der Medizin zu analysieren, müssen wir daher zwei wichtige Fragen beantworten: 1) Wie kann man mit dem Fluch der Dimensionalität und der Begrenztheit der annotierten Daten umgehen? 2) Wie kann die Transparenz von DL-Modellen durch Interpretierbarkeit verbessert werden, da dies potenziell zu einem besseren Verständnis der Daten und des eingesetzten Modells führt?
Diese Arbeit befasst sich mit diesen Herausforderungen in strukturierten Hochdurchsatzdaten und hochdimensionalen Bildgebungsdatenmodalitäten. Wir beginnen unsere Studie über strukturierte Hochdurchsatzdaten mit der Anwendung der Proteomik-Datenanalyse. Wir erlernen die Datenrepräsentation auf robuste Weise und extrahieren die medizinisch relevanten Informationen mithilfe von DL-Techniken. Wir entwickeln neuartige Datenanalysen, die auf dem basieren, was das DL-Modell durch die Interpretation seiner Vorhersagen lernen kann. Diese Informationen ermöglichen einen Einblick in die Datenmuster und die Entdeckung von Unterscheidungsmerkmalen. Wir rechtfertigen auch die Zuverlässigkeit der Modellinterpretation durch umfassende quantitative Bewertungen. Wir zeigen, dass die richtige Kombination von DL-Techniken in Verbindung mit Interpretationsstrategien, die ein tiefgreifendes Verständnis der Modellentscheidungen ermöglichen, zu einem zuverlässigen klinischen Entscheidungsunterstützungssystem führen kann.
Außerdem untersuchen wir DL-Techniken für hochdimensionale Bilddaten. Im Gegensatz zu strukturierten Daten, bei denen die gewünschten Merkmale mit geringen Abweichungen auftreten, können die interessierenden Regionen auf medizinischen Bildern an verschiedenen Datenpunkten mit einer großen Abweichung auftreten. Daher haben wir unsere Bildanalyse auf überwachte convolutional neural network (CNN) aufgebaut, die mit großen Abweichungen umgehen können.
Wir untersuchen verschiedene CNN-Architekturen und vergleichen ihre Stärken. Schließlich haben wir eine robuste Pipeline für heterogene Bilddaten mit der anspruchsvollen Anwendung der Erkennung und Identifizierung menschlicher Wirbel entwickelt.
Um mit der Datenknappheit umzugehen, zeigen wir, wie verschiedene Techniken, einschließlich Transferlernen, Datenerweiterung, Human-in-the-Loop und synthetische Generierung von Daten im medizinischen Umfeld, die Generalisierung verbessern
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Get PDF“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Get PDFWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
Dispelling the Myths Behind First-author Citation Counts
Get PDFWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
koamabayili/VECTRON-author-checklist: VECTRON author checklist
No full textWe have done our best to complete the author checklist relating to the use of animals in the hut study. Note that the objective for the hut study was to evaluate the IRS treatment applications for residual efficacy against Anopheles mosquitoes, including the local An. coluzzii mosquito population. Cows were only used to attract mosquitoes into the huts and no tests were carried out directly on the cows. The author checklist is intended for use with studies where experiments are carried out on animals, which is why we have had such difficulty in completing this for the hut study, as many of the questions do not relate to how the cows were used
Author-wise bibliometric analysis based on entropy.
No full textAuthor-wise bibliometric analysis based on entropy.</p
- …
