• Zum Inhalt springen
  • Zur Navigation springen
  • Zum Seitenende springen
Organisationsmenü öffnen Organisationsmenü schließen
Friedrich-Alexander-Universität Lehrstuhl für Wirtschaftsinformatik
  • FAUZur zentralen FAU Website
  1. Friedrich-Alexander-Universität
  2. Fachbereich Wirtschafts- und Sozialwissenschaften
  • Mein Campus
  • UnivIS
  • StudOn
  • Lageplan
  1. Friedrich-Alexander-Universität
  2. Fachbereich Wirtschafts- und Sozialwissenschaften
Friedrich-Alexander-Universität Lehrstuhl für Wirtschaftsinformatik
Menu Menu schließen
  • Home
  • Lehre
    • Bachelor
    • Master
    • Learning Agreements
    Portal Lehre
  • Forschung
    • Projekte
    • Veröffentlichungen
    • Förderer & Partner
    Portal Forschung
  • Team
  • Abschlussarbeiten
  1. Startseite
  2. Abschlussarbeiten
  3. Offene Abschlussarbeiten

Offene Abschlussarbeiten

Bereichsnavigation: Abschlussarbeiten
  • Allgemeine Informationen und Vorlagen
  • Offene Abschlussarbeiten

Offene Abschlussarbeiten

 

Thema Betreuer Typ Partner
Evaluation of Computer Vision Architectures Using Explainable AI (XAI) Methods

Computer Vision is a subfield of Machine Learning (ML) that typically uses Convolutional Neural Networks (CNNs) to process Images. Instead of using CNNs, Google lately suggested new architectures (see e.g., Transformer and Mixer) to deal with spatial data. Recently, the Vision Transformer (ViT) has emerged as a competitive alternative to Convolutional Neural Networks (CNN), which are currently the state of the art in Computer Vision and are therefore widely used in various image recognition tasks.

The goal of this thesis is to compare a CNN (e.g. ResNet50) and a ViT using XAI methods, like GradCam. More details will follow in consultation.
Important: Very good knowledge in statistics is required to work on this thesis! Expected starting date is as soon as possible.

Dilara Yesilbas Bachelor/Master –
Evaluating Architectures for Computer Vision

Computer Vision is a subfield of Machine Learning (ML) that typically uses Convolutional Neural Networks (CNNs) to process Images. Instead of using CNNs, Google lately suggested new architectures (see e.g., Transformer and Mixer) to deal with spatial data. Recently, the Vision Transformer (ViT) has emerged as a competitive alternative to Convolutional Neural Networks (CNN), which are currently the state of the art in Computer Vision and are therefore widely used in various image recognition tasks.

The goal of this thesis is to compare a CNN (e.g. ResNet50), ViT, (and MLPMixer). More details will follow in consultation.
Important: Very good knowledge in statistics is required to work on this thesis! Expected starting date is as soon as possible.

Dilara Yesilbas Bachelor/Master –
State-of-the-Art in Natural Language Processing 

Natural Language Processing (NLP) is a subfield of Machine Learning (ML) that deals with understanding and generating text. The aim of the studies is to summarize and distinguish the state-of-art neural architectures, public benchmark dataset etc. on one of the pre-defined tasks.

  • Text Classification (Can ML categorize texts?)
  • Sentiment Analysis (Can ML assess feelings?)
  • Entity Recognition (Can ML extract named entities?)
  • Paraphrase Identification (Can ML recognize duplicates?)
  • Question Answering (Can ML answer questions?)
  • Textual Entailment (Can ML deal with hypothesis?)
  • Document Intelligence (Can ML process scans?)
  • Text Summarization (Can ML shorten texts?)

Depending on your prior knowledge, the studies may take the form of literature reviews or experiments (code snippets available). Note that a NVIDIA GPU is recommended for training and executing machine learning models, esp. Transformer-based architectures such as BERT.

 

Stefan Arnold Bachelor/Master –
Evaluating Architectures for Computer Vision w.r.t. Spurious Correlations in Images: A Fairness Perspective

Computer Vision is a subfield of Machine Learning (ML) that typically uses Convolutional Neural Networks (CNNs) to process Images. Instead of using CNNs, Google recently suggested new archtectures (see e.g., Transformer and Mixer) to deal with spatial data. While the top-line generalization accuracy has been comparatevly to CNNs, the fairness (measured as the parity of accuracy) among subgropus remains an open question.

Using the publicly available CelebA dataset, the goal of this thesis is to compare ResNet18 (representative for CNNs), ViT, and MLPMixer in terms of their accuracy within ethnical subgroupus (i.e., Gender and Age). See Sagawa et al. (2020) for the methodological setup (also code snippets available upon request) and Hooker et al. (2020) for a rigorous assessment of algorithmic bias.

NVIDIA GPU or Google Colab Prenium recommended.

Stefan Arnold Master –
Real-Time 3D Visualization of Queues with ML-Based Prediction of Item Processing for a Product Information Management System (VERGEBEN)
We are all familiar with queues and waiting times—but none of us like them. However, queues are an important concept for companies, where “queuing” as a service is emerging as an important Business model. The thesis should involve in developing a prototype system for predicting and visualizing distributed product information Queues. It contributes to the ongoing academic discourse regarding the applications of artificial intelligence to business and the practical development of prediction and visualization solutions.The prototype can be found on GitHub https://github.com/Sultanow/dc_cubesVoraussetzungen:
  • Diese Arbeit richtet sich vor allem an Studierende der Studiengänge Wirtschaftsinformatik mit praktischen Erfahrungen in gängigen Programmiersprachen (Insbesondere Python) und Tools.
Mark Kram Bachelor/Master Capgemini
Machinelles Lernen in der statischen Codeanalyse (VERGEBEN)
Die Assoziationsanalyse ist ein maschinelles Lernverfahren aus dem Bereich des unüberwachten Lernens, das ursprünglich durch Anwendungen aus dem Handelsumfeld motiviert ist. Ihr liegt das Prinzip zugrunde, bedingte Wahrscheinlichkeiten des Erwerbs eines Produkts A unter der Bedingung des Erwerbs eines anderen Produkts B zu ermitteln. Das Verfahren wird hier auf die statische Codeanalyse (SCA) zur Identifizierung potenzieller Fehler eines missionskritischen Großsystems im öffentlichen Sektor angewendet. Verzögert sich dort die Fehlerbereinigung, kann es je nach Schwere des Fehlers teuer werden – ein Hotfix ist dann unabdingbar. Im Rahmen der Abschlussarbeit soll ein Algorithmus und Werkzeug optimiert werden, dass sich die Assoziationsanalyse für SCA in einem derart kritischen Umfeld zunutze macht.Machine Learning is often associated with predictive analytics, for example with the prediction of buying and termination behavior, with maintenance times or the lifespan of parts, tools or products. However, Machine Learning can also serve other purposes such as identifying potential errors in a mission-critical large-scale IT process of the public sector. A delay of troubleshooting can be expensive depending on the error’s severity – a hotfix may become essential. The Thesis should optimize an algorithm and tool that uses unsupervised machine learning techniques for SCA in such a critical Environment.Voraussetzungen:
  • Diese Arbeit richtet sich vor allem an Studierende der Studiengänge Wirtschaftsinformatik mit praktischen Erfahrungen in gängigen Programmiersprachen (Insbesondere Python) und Tools.

 

Mark Kram Bachelor/Master Capgemini

Relational Text Mining am Beispiel der Startup-Szene

„Relation Extraction is the task of predicting attributes and relations for entities in a sentence. For example, given a sentence “Barack Obama was born in Honolulu, Hawaii.”, a relation classifier aims at predicting the relation of “bornInCity”. Relation Extraction is the key component for building relation knowledge graphs, and it is of crucial significance to natural language processing applications such as structured search, sentiment analysis, question answering, and summarization“ (arXiv:1707.08866).

Ziel der Arbeit ist es, anhand eines konkreten Anwendungsbeispiels aus der Domäne der Startupbranche, einen NLP-basierten Algorithmus oder Teile davon prototypisch zu implementieren. Hierbei sollen als Datengrundlage News-Artikel der Startupbranche als unstrukturierte Textdokumente genommen werden.

Voraussetzungen:

  • Grundkenntnisse in Python oder anderen Programmiersprachen
  • Grundkenntnisse oder Interesse an ML-Modellen, insbesondere im Bereich Natural Language Processing

 

Mark Kram Bachelor/Master –

Comparative Analysis of Low Code Machine Learning Tools (NOT AVAILABLE)

„A low-code development platform (LCDP) provides a development environment used to create application software through a graphical user interface instead of traditional hand-coded computer programming. A low-coded platform may produce entirely operational applications, or require additional coding for specific situations. Low-code development platforms reduce the amount of traditional hand coding, enabling accelerated delivery of business applications. A common benefit is that a wider range of people can contribute to the application’s development—not only those with coding skills.“ – Wikipedia

Low- and No-Code platforms are becoming increasingly popular for machine learning (ML) and data science use cases, as they lower the skill requirements to implement machine learning based solutions.The goal of this thesis is to make an empirical comparison of different Low- and No-Code tools for ML use cases. The focus can be specified on certain areas of ML, such as NLP.

 

Further reading:

https://medium.com/aixdesign/26-no-code-low-code-ml-tools-to-check-out-8b3414eaa489

https://towardsdatascience.com/top-8-no-code-machine-learning-platforms-you-should-use-in-2020-1d1801300dd0

https://www.analyticsinsight.net/top-low-code-and-no-code-ml-platforms-for-beginners/

 

Requirements:

  • Experience or interess in programming and machine learning

 

Mark Kram Bachelor/Master
Graph Neural Networks meet Reinforcement Learning – A Systematic Literature Review (NOT AVAILABLE)


Graph Neural Networks (GNNs) have gained interest in the past years, as they provide a suitable neural networks based technique to process graph and network data. Typical graph data are social networks, used in platforms like Twitter and Facebook.Recently, there have been efforts to combine GNNs with Reinforcement Learning with promising results. Exemplary use cases are fraud detection and molecular property prediction (E.g. for medicinal drug discovery).The aim of this work is to provide a comprehensive systematic literature analysis for GNNs in combination with Reinforcement Learning.

 

Mark Kram Bachelor/Master
Towards Malicious Smart Contract Detection with Graph Neural Networks (NOT AVAILABLE)

 

Smart Contracts are simply programs stored on a blockchain that run when predetermined conditions are met. They typically are used to automate the execution of an agreement so that all participants can be immediately certain of the outcome, without any intermediary’s involvement or time loss.However, a malicious user can deploy vulnerable smart contracts to breach the blockchain data. Also, authors with lacking coding skills may write contracts that they do not realize are vulnerable, just as a poorly written law might have unintended loopholes in it.In the past years, Graph Neural Networks (GNNs) have been applied in the field of anomaly and fraud detection with promising results. The reason to this is that many fraud cases can be formulated as a graph or network problem, e.g. fake news in a social network, finance fraud in a financial transaction network or, in this case, vulrerable smart contracts in the blockchain graph. It therefore seems suitable to apply GNNs to the Smart Contract Anomaly Problem.

The aim of this work is therefore to investigate, how GNNs can be applied to detect vulnerable/malicious smart contracts.

 

Literature:

https://www.semanticscholar.org/paper/A-Comprehensive-Survey-on-Graph-Anomaly-Detection-Ma-Wu/5e4c5f7506a1bb8aab0a9d5168674766eb54a91f

 

Prerequisites:

– Knowledge about Blockchain and smart contract technology.

– Knowledge and experience in deep learning techniques.

– Knowledge or skills in programming languages.

Mark Kram (Bachelor)/Master

Neural Recommendation System for Academic Contents – A Graph Neural Network Based Approach at FAU

Recommendation Systems have been successfuly used in many areas, like marketing and E-commerce. A reason is that this technology tackles the problem of information overload, which often occurs in decision processes.

For this thesis, the aim is to implement a prototype for a recommendation algorithm for the academic content at the FAU with the usage of Graph Neural Networks. These type of deep learning technique are used in many state of the art recommendation approaches (e.g. recommendations at pinterest: shorturl.at/hFNP1).  Data (Already preprocessed) and a first graph structure will be provided.

Prerequisites:

  • Experience in programming languages like python and machine learning frameworks
  • Interess or experience in graph based approaches
Mark Kram Bachelor/Master

 

Lehrstuhl für Wirtschaftsinformatik,
insb. IT-Management

Lange Gasse 20
90403 Nürnberg
  • Institut für WIN
  • Impressum
  • Datenschutz
  • Barrierefreiheit
Nach oben