TU Wien:Generative AI VU (Heitzinger)
- Generative AI VU (Heitzinger) (TU Wien, 1 Material)
- Generative AI VU (Neidhardt) (TU Wien, 3 Materialien)
- Generative AI in Planning and Architecture concepts rendering and post-rendering VO (Korolova) (TU Wien, 0 Materialien)
Daten[Bearbeiten | Quelltext bearbeiten]
| Vortragende | Mohammad Mahdi Azarbeik• Georg Gottlob• Clemens Heitzinger• Bernhard Krüpl-Sypien• Julia Neidhardt• Emanuel Sallinger |
|---|---|
| ECTS | 3,0 |
| Letzte Abhaltung | 2024W |
| Sprache | English |
| Mattermost | generative-ai • Register • Mattermost-Infos |
| Links | tiss:194154 |
Inhalt[Bearbeiten | Quelltext bearbeiten]
2024WS:
This course covers the following topics:
- Basics of Language Modelling
- Prompt Engineering
- Deep Dive Transformers
- Deep Dive PPO & RLHF
- Recommender Systems in Generative AI
- Knowledge Graphs in Generative AI
- Applied Generative AI
- Ethics of Generetive AI
- Diffusion models and research outlook
Ablauf[Bearbeiten | Quelltext bearbeiten]
2024WS:
- Weekly Lectures (attendance is not mandatory)
- 6 Homework Assignments
- 2 Multiple-Choice Tests
Benötigte/Empfehlenswerte Vorkenntnisse[Bearbeiten | Quelltext bearbeiten]
noch offen
Vortrag[Bearbeiten | Quelltext bearbeiten]
noch offen
Übungen[Bearbeiten | Quelltext bearbeiten]
Almost all the homework assignments are to be done in Python. They are designed to be solved quickly with just a few lines of code.
Prüfung, Benotung[Bearbeiten | Quelltext bearbeiten]
Calculation of grades: homework assignments 50%, each test 25%. To pass the course one needs at least 50% of the total test points and 50% of the total homework assignments points.
Dauer der Zeugnisausstellung[Bearbeiten | Quelltext bearbeiten]
2024WS: Grades were posted on 21.02.2025
Zeitaufwand[Bearbeiten | Quelltext bearbeiten]
2024WS: I spent 19 hours in total. This includes 4.5 hours preparing for Test 1, 6 hours preparing for Test 2, and the remaining time was spent working on homework assignments, which typically took 1 to 2 hours each. Grade obtained: 2.
Unterlagen[Bearbeiten | Quelltext bearbeiten]
noch offen
Tipps[Bearbeiten | Quelltext bearbeiten]
noch offen
Highlights / Lob[Bearbeiten | Quelltext bearbeiten]
No group work
Verbesserungsvorschläge / Kritik[Bearbeiten | Quelltext bearbeiten]
2024WS: "Too many cooks spoil the soup." It feels like the course was meant to be improved by bringing in domain experts for most topics, but in the end, it became too confusing and disorganized. The lack of a clear common thread made everything feel asynchronous and somewhat disconnected. This was the first time in my studies that I genuinely felt the course would have been better if just one person had taught the whole thing.