TU Wien:Recommender Systems VU (Neidhardt)

Daten

Zuordnungen

Masterstudium Data Science

Modul MLS/CO - Machine Learning and Statistics - Core (Gebundenes Wahlfach)

Inhalt

noch offen, bitte nicht von tiss oder Homepage kopieren, sondern aus Studierendensicht beschreiben.

Ablauf

• 4 short exercises on JupyterHub (20 points)
• 1 group project with 5-8 persons (30 points)
• 1 exam, 60 minutes (50 points)

2026S

The lectures for Recommender Systems and User Modeling (6 ECTS) and Recommender Systems (3 ECTS) were held together. Because there were more students than the capacity of the lecture room, it was decided to teach in a hybrid form (in class lecture + live stream on Lecture Tube). The lecture was only live streamed but not recorded. Presence was not mandatory during the lecture.

The slides were provided after the lecture. Students noted that it would be helpful to have access to the slides before the lecture, but the course team decided to keep their approach.

Benötigte/Empfehlenswerte Vorkenntnisse

• Exercises: Basic python programming
• Project: Basic python programming, Neural Network and Git knowledge very useful

Vortrag

noch offen

Übungen

2024S

Individual assignments 1 to 4 proceeded as before. This time, the group assignment involved solving the RecSys Challenge related to recommending Danish news articles (see https://recsys.eb.dk/).

Fundamentally, the task (in groups of five) was actually interesting and enjoyable - however, nearly every circumstance prevented a thorough engagement with the content. Ultimately, this led to an incredible increase in time expenditure (60+ hours per person), which I have not previously encountered at the TU. Points of frustration included:

• A poorly documented, buggy code framework.
• Significant limitations of the environment intended for training - consistent various error messages and hardware restrictions.
• Unclear specifications regarding submission and execution.

Compiling a functioning submission, including data, took an exceedingly long time. While well-intentioned, the exercise turned out to be excessive! Hopefully, this will not occur again. Additionally, the Danish newspaper involved is quite a low-quality publication similar to "Bild" in Germany - though this is, of course, irrelevant here.

To this time, grading has not be done... let's see but really difficult to do a fair rating with all hurdles and team issues. Really not a comfy situation.

Prüfung, Benotung

Dauer der Zeugnisausstellung

noch offen

Zeitaufwand

noch offen

Unterlagen

noch offen

• Materialien

Tipps

noch offen

Highlights / Lob

noch offen

Verbesserungsvorschläge / Kritik

noch offen

Materialien

Neues Material hinzufügen

2

• 2024S recommender systems open.docx (details)

@

• @sueszli's exercise solutions (details)

E

• Exam 2022-06-28

L

• Lernkartei-recommender-systems.pdf (details)

S

• Sueszli's summary + altprüfungen (details)