Modulcode:	infMPGAI-01a
Englische Bezeichnung:	Master Project - Generative AI
Modulverantwortliche(r):	Prof. Dr. Sören Pirk
Turnus:	unregelmäßig (WS24/25)
Präsenzzeiten:	4PÜ
ECTS:	10
Workload:	300 Std. Projektarbeit, davon 60 Std. betreut
Dauer:	ein Semester
Modulkategorien:	MSc-Inf-Proj (MSc Inf (21))
Lehrsprache:	Englisch
Voraussetzungen:

Kurzfassung:

This master project provides the opportunity to apply and advance the theoretical knowledge of training deep neural networks and generative AI approaches to a concrete practical problem. At the start of the semester students will pick concrete projects, for which they work together through the semester in small teams to achieve the goals.

Lernziele:

Students will learn to work on a complex problem with a focus on machine learning and generative AI. This includes the use of the required software libraries (Tensorflow, Pytorch, Python) and the required frameworks for training neural networks. A key goal of the project is to either use existing state-of-the-art neural network architectures, to develop novel architectures, or to apply AI models to applications toward medical diagnostics, manufacturing, and autonomous agents. Specifically, the projects are setup to leverage deep learning models for image generation, 3D shape processing, or learning models for time-series data. Students will also practice to work in a team, to analyze the requirements for a practical project, and to plan the steps to achieve the goals.

Lehrinhalte:

Advanced concepts of Generative AI, including Generative Adversarial Networks, Diffusion Models, Domain Adaptation and Style Transfer, Neural Radiance Fields (NeRFs), and Synthetic Data Generation. The project will be implemented in Physthon or in conjunction with state-of-the-art rendering engines (e.g. Unreal Engine 5 with C++).

Weitere Voraussetzungen:

This master project targets students that have ideally completed the class NNDL or GenAI or those who have equivalent knowledge such that they understand advanced concepts of modeling and rendering.

Prüfungsleistung:

Presentations including demonstration, report and the completed software system (incl. documentation).

Lehr- und Lernmethoden:

Students will work on an individually defined project.

Verwendbarkeit:

Literatur:

• Deep Learning, Ian Goodfellow and Yoshua Bengio and Aaron Courville, 2016, MIT Press
• Generative AI - Teaching Machines to Paint, Write, Compose and Play, David Foster, 2019, O'Reily

Verweise:

Kommentar:

There are only a few slots available in this course. Please contact Sören Pirk directly by email (sp@informatik.uni-kiel.de) in case you are interested. Approval is mandatory before registering for the class. The first meeting for the project will be announced at the beginning of the semester.