Labs and Seminars

Two types of practical classes will be proposed for the course (alternating):

  • labs, which combine a tutorial on tools and a discussion of the homework assignment. Students will implement selected methods discussed in the course and experiment with them at home
  • theoretical seminars in which students will discuss solutions of theoretical assignments (posted 1 week before the class)

Schedule

(contents will be updated)

Date Topic Teacher Reading
18.02.2026 Lab 1: Preparations, Double Descent + MLP AS, VS NNTD lecture 1 (except 2.3, 5)
25.02.2026 Seminar 1 JH
04.03.2026 Lab 2: CNN Finetuning & Visualization + AY
11.03.2026 Seminar 2 AS
18.03.2026 Lab 3: From Scratch: Initialization & regularization TA
25.03.2026 Seminar 3 JH
01.04.2026 Lab 4: Transformers BP
08.04.2026 Seminar 4 AS
15.04.2026 Lab 5: Metric Learning KZ/PS
22.04.2026 Seminar 5 JH
29.04.2026 Lab 6: Graph Neural Networks VS
06.05.2026 Seminar 6 AS
13.05.2026 — Rector's day —
20.05.2026 Lab7: TBA TA+BP

Seminars

The seminar assignments are published 1 week in advance before the seminar. You are expected to prepare for it at home. We discuss the problems and solutions in the class. You are not required to submit you solutions, but if you solved a problem you will be invited to present it in the class. Seminars are not scored by points but they are important for gaining technical understanding, which will be finally evaluated in the written exam.

Examples of problems with solutions: examples.pdf (to be updated)

Labs

The solutions of the practical labs have to be submitted using the upload system

  • Your task will be to program a solution of the assigned problems. You have to hand out your code and a report. The report has to contain only answers to the assignments (nothing else).
  • The programming language is Python/PyTorch.
  • The deadline for submitting your solutions will be 2 weeks after the date of assignment. This is a soft deadline, if you still submit within 3 weeks you get a deduction of 3 points from the maximum 10. End of the week 3 is a hard deadline.
  • Not submitting a lab is equivalent to getting 0 points. You need at least of 50% of total lab points to pass.

Submission Regulations

We expect you to submit:

  1. Python source code without redundancies (e.g. series of failed attempts).
  2. Report in pdf / printed Jupyter notebook with inline results

Sharing code that is not a required part of the assignment is permitted, for example additional visualization code or test cases for debugging.

courses/bev033dla/labs/start.txt · Last modified: 2026/02/11 16:01 by shekhole