|
|
Vorlesungsverzeichnis >> Fakultät Wirtschaftsinformatik und Angewandte Informatik >>
|
xAI-Proj-B: Bachelorprojekt Erklärbares Maschinelles Lernen (xAI-Proj-B)- Dozentinnen/Dozenten
- Sebastian Dörrich, M.Sc., Prof. Dr. Christian Ledig
- Angaben
- Übung
Rein Präsenz
4,00 SWS, Unterrichtssprache Englisch
Zeit und Ort: Do 14:00 - 18:00, WE5/03.004
- Englischsprachige Informationen:
- Title:
- xAI-Proj-B: Bachelor Project Explainable Machine Learning
- Prerequisites
- none
- Contents
- The popularity of artificial intelligence (AI) has grown exponentially in recent years and has been the talk of the town ever since the release of ChatGPT at the end of 2022. AI systems can solve complex problems, understand and generate human language, control robots, or even compose poetry. The current driving force behind AI and destined "to lead the field of AI to its next spring" is commonly known as the area of "Deep Learning (DL)". Deep learning involves the use of artificial neural networks to learn from vast amounts of data and improve performance through experience. By training these neural networks on large datasets, deep learning models can recognize patterns and features in data that may be too complex for traditional machine learning algorithms. This fact makes deep learning a powerful tool in today's society and comes hand-in-hand with a high demand for professionals who can develop and implement these algorithms.
This project aims to provide a first dive into the field of deep learning by presenting a comprehensive view of what it takes to be a DL engineer in the real world. For this, we will mimic their routine in approaching a new task over the course of a semester. Working in teams of 4, the students will nurture their own DL model on its path from initialization and adaption to simple problems all the way up to the application in a real-world scenario. This will include tasks such as researching, data collection, implementation, experimenting, and of course testing the models.
More specifically throughout the project, you will develop a neural network for the classification of street numbers in images. For this, each team will first develop a neural network classifier for synthetic, simplistic data (images of handwritten digits) to validate that your idea is conceptually working. For the implementation, we will use Python and the prominent deep learning framework PyTorch. Afterward, we will work together to create a one-of-a-kind dataset of images of street numbers by doing some "field work" ;). Each team will further adapt its model for this new dataset to enable its usage for an actual real-world problem. As a last step, you will present your results to your fellow researcher colleagues, participate in interesting discussions, and hand in a technical project report describing all your ideas, approaches and results. Oh and to make the project even more thrilling, you will be able to compete with all the other teams through a real-life competition for the first prize!
Further information:
All further information regarding the project will be provided at the first event on Thursday, 20th of April 2023 at 2:00-6:00 pm in room WE5/03.004.
- Literature
- If you are bored and have some free time, feel free to check out some of the following Tutorials about Deep Learning / PyTorch in advance
Deep Learning Crash Course for Beginners: https://www.youtube.com/watch?v=VyWAvY2CF9c
Pytorch - Deep Learning w/ Python: https://www.youtube.com/playlist?list=PLQVvvaa0QuDdeMyHEYc0gxFpYwHY2Qfdh
Deep Learning With PyTorch - Full Course: https://www.youtube.com/watch?v=c36lUUr864M
Beginner Deep Learning Tutorial in PyTorch How to Make a Convolutional Neural Network Tutorial 1: https://www.youtube.com/watch?v=H69j69FFMV0
...
- Zusätzliche Informationen
- Erwartete Teilnehmerzahl: 20
- Institution: Lehrstuhl für Erklärbares Maschinelles Lernen
Hinweis für Web-Redakteure:
Wenn Sie auf Ihren Webseiten einen Link zu dieser Lehrveranstaltung setzen möchten, verwenden Sie bitte einen der folgenden Links:Link zur eigenständigen Verwendung Link zur Verwendung in Typo3
|
 |
|
|
UnivIS ist ein Produkt der Config eG, Buckenhof |
|
|