TU Wien:AI/ML in the Era of Climate Change VU (Brandic)

Aus VoWi
Zur Navigation springen Zur Suche springen

Daten[Bearbeiten | Quelltext bearbeiten]

Vortragende Ivona BrandicShashikant Shankar Ilager
Letzte Abhaltung 2023WS
Sprache English
Mattermost ai-ml-in-the-era-of-climate-changeRegisterMattermost-Infos
Links tiss:194125
Master Data Science Wahlmodul MLS/EX - Machine Learning and Statistics - Extension

Inhalt[Bearbeiten | Quelltext bearbeiten]

noch offen, bitte nicht von TISS/u:find oder Homepage kopieren, sondern aus Studierendensicht beschreiben.

Ablauf[Bearbeiten | Quelltext bearbeiten]

2 Exercises and an Oral Exam

Benötigte/Empfehlenswerte Vorkenntnisse[Bearbeiten | Quelltext bearbeiten]

For the exercises you should be familiar with neural network frameworks and machine learning in general, as it is required for the exercise but not covered in the lectures

Vortrag[Bearbeiten | Quelltext bearbeiten]

noch offen

Übungen[Bearbeiten | Quelltext bearbeiten]

WS2022: 2 Exercises in teams of 3:

  • Exercise 1: Take a pretrained model for the COCO dataset (object detection in images) and use different kinds of quantization on the model using Tensorflow Lite. Observe how quantization changes things like accuracy, inference time and model size. Since object detection isn't such a simple task as regression or classification, just getting the pretrained model running actually took our group much more time than the quantization that this exercise was aiming at.
  • Exercise 2: Using the LamaH dataset (meterological data), "forecast" precipitation (rain) using different machine learning models ranging from linear models to neural networks. There was no information on what was meant with "forecast" i.e, what variables to use, how many past days to take into account or how many days into the future to predict. In the following exercise presentation it seemed that pretty much anything you did (even just doing regression instead of forecasting) was fine.

Exercises were followed by 10 minute online presentation by each group.

Prüfung, Benotung[Bearbeiten | Quelltext bearbeiten]

WS2022: Oral Exams in groups of 4 with 30 minutes per group. Each group was assigned a paper and each student in the group got 3 questions:

  • 1 question from the contents of the lecture
  • 1 question from the exercise
  • 1 question from the paper that the group was assigned

If you do the math you'll see that a student has about 2,5 minutes for each question. Of course this also includes the time needed to ask to question, so actually it's closer to 2 minutes. Since students didn't really have time to think much about the questions everybody ended up just blurting out some fancy sounding keywords and trying to fit them into a sentence somehow, especially for the question from the lecture contents. This wan't really an issue for the grade, as I think literally any answer that could somehow be interpreted as trying to say the correct thing was accepted.

Dauer der Zeugnisausstellung[Bearbeiten | Quelltext bearbeiten]

WS2022: ~3 weeks after oral exam

Zeitaufwand[Bearbeiten | Quelltext bearbeiten]

Less than 4 ECTs

Unterlagen[Bearbeiten | Quelltext bearbeiten]

noch offen

Tipps[Bearbeiten | Quelltext bearbeiten]

noch offen

Highlights / Lob[Bearbeiten | Quelltext bearbeiten]

noch offen

Verbesserungsvorschläge / Kritik[Bearbeiten | Quelltext bearbeiten]

noch offen