Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ edoc-Server. Open-Ac...arrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
https://dx.doi.org/10.18452/34...
Doctoral thesis . 2025
Data sources: Datacite
DBLP
Doctoral thesis
Data sources: DBLP
versions View all 3 versions
addClaim

Deep Learning for Sequential Data

Authors: Zinovyeva, Elizaveta;

Deep Learning for Sequential Data

Abstract

Deep Learning (DL) ist ein Teilgebiet des maschinellen Lernens (ML), das auf neuronalen Netzen basiert, die auf verschiedenen Abstraktionsebenen lernen und dabei traditionelle ML Methoden in einer Vielzahl von Aufgaben häufig übertreffen. Diese Dissertation untersucht verschiedene Geschäftsaufgaben mit sequenziellen Daten und analysiert die DL Eignung. Zuerst stellen wir den DL Begriff und die Taxonomie der sequenziellen Datenforschung vor. Die ermittelte Taxonomie zeigt die Dimensionen auf, entlang derer unsere Analyse von DL für die sequenzielle Datenforschung durchgeführt wird. Kapitel 2 untersucht die Machbarkeit und Eignung der automatischen Inhaltsüberwachung und deren Erklärbarkeit. Wir konsolidieren die Arbeiten zur Erkennung antisozialen Verhaltens und identifizieren DL- sowie nicht-DL-basierter Algorithmuskandidaten und betonen die Bedeutung der Interpretierbarkeit. Kapitel 3 widmet sich der jüngsten Popularität von Smart Contracts (SC), Softwareprogrammen, die auf einer Blockchain basieren und bestimmte Aktionen ausführen können. Wir befassen uns mit den rechtlichen, rechnerischen und statistischen Aspekten von SCs und untersuchen, ob reale Anwendungen mit den vorgeschlagenen Medienanwendungsfällen korrelieren und potenziell die Wirtschaft revolutionieren können. DL-Methoden dienen als Instrument zum Verständnis des SCs-Ökosystems. Kapitel 4 stellt das Code-Zusammenfassungs (CZ) ‚Algorithm Choice Framework‘ für den Wissenschaftsbereich vor, das Wissenschaftler bei der Entwicklung ihres CZ-Systems unterstützt. Die Anwendung des Frameworks wird anschließend am Anwendungsfall von Quantlet.com veranschaulicht. Abschließend untersucht Kapitel 5 die Überwachung der CZ-Systeme. Das Kapitel schlägt das Meta-Learning-Modul vor, das die Überwachung, Modellauswahl und Ensemblebildung von CZ-Modellen mithilfe einer Mischung von Experten ermöglicht. Abschließend wird erörtert, wie das Modul eine prädiktive Perspektive auf die Interpretierbarkeit bieten kann.

Deep Learning (DL) is a subset of machine learning (ML) based on neural networks that learn at different abstraction levels, frequently outperforming traditional ML methods in a variety of tasks. This dissertation investigates various business tasks involving sequential data and analyzes the DL methods suitability. First, we introduce the notion of deep learning and the taxonomy of sequential data research. The identified taxonomy showcases the dimensions along which our holistic analysis of DL for sequential data research is conducted. The second chapter investigates the feasibility and suitability of automatic content monitoring and the explainability of such systems. Here, we consolidate the work on antisocial behavior detection and identify algorithm candidates, including DL and non-DL-based methodologies, and highlight the importance of the interpretability. The third chapter is dedicated to the recent popularity of Smart Contracts (SC), the software programs that live on a blockchain and can perform specific actions. Here, we dive into the legal, computational, and statistical aspects of SCs, identifying whether real-world applications correlate with the suggested media use cases and can potentially revolutionize the economy. DL methods serve as a supporting tool for understanding the SCs ecosystem. The fourth chapter proposes the code summarization Algorithm Choice Framework for the science domain, supporting scientists in designing their code summarization system. The framework’s usage is then exemplified in the real-world use case of Quantlet.com. Finally, chapter 5 investigates monitoring the code summarization systems. The chapter proposes the Meta-Learning Module that enables code summarization model monitoring, model selection, and ensembling using a mixture of experts. Finally, it discusses how the module can give a predictive perspective on interpretability.

Country
Germany
Related Organizations
Keywords

QH 700, ddc:000, ddc:330, Code, QH 232, 330 Wirtschaft, QH 234, ST 306, Sequential Data, Text, Machine Learning, Deep Learning, 000 Informatik, Informationswissenschaft, allgemeine Werke, ST 302, MS 4855, Maschinelles Lernen, Sequentielle Daten

  • BIP!
    Impact byBIP!
    selected citations
    These citations are derived from selected sources.
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
Powered by OpenAIRE graph
Found an issue? Give us feedback
selected citations
These citations are derived from selected sources.
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average
Green