Seminar in Deep Neural Networks (FS 2022)
The Seminar is now concluded.
As a seminar participant, you are asked to attend all the talks and make a presentation.
Here is the link to the seminar's Moodle course. You can use Moodle to ask questions or start discussions.
Here is the link to the seminar's Polybox repository. You should have received the password by email. You should upload the slides for your talk the day before it is scheduled.
Disclaimer: This is a seminar, we will focus on reasearch and skip most of the basics. If you feel like you cannot follow the discussions we invite you to check out this lecture or other lectures on deep learning. We also recommend Chollet, Francois: Deep learning with Python. 2nd Edition. Simon and Schuster, 2021.
Presentation & Discussion
Your presentation should be in English. The presentation should last exactly 35 minutes, followed by a discussion. The seminar will be exciting when the talks and discussions are exciting.
Here are some guides and guidelines on scholarly presentations: by Roger Wattenhofer, by Garr Reynolds, and by Cheryl Gore-Felton. As is customary in academia, all work not belonging to the author (including figures, explanations, examples, or equations) must be properly referenced.
We expect the presentation to motivate a lively discussion. We encourage discussions during and after the presentations as a main objective of this seminar. It may help discussions if you also try to be critical about the presented work. Participating in Moodle discussions also counts towards the discussion participation.
We plan to hold this year's seminar in person. If you find yourself unable to participate (say due to an isolation requirement), get in touch with your mentor at first instance. You might then be asked record a video of your presentation and upload it to polybox the day before your seminar presentation date.
Your grade will depend on your presentation for the most part. In addition, we also grade how actively you participate in the discussions, both in person and on Moodle, throughout the semester.
Presentation TimelineWe established the following rules to ensure a high quality of the talks:
- As soon as you know who your mentor is: ask them for a literature assignment in case they haven't provided it already.
- At least 4 weeks before your talk: first meeting with your mentor where you discuss the structure of the talk.
- 4 to 1 week before your talk: you meet with your mentor as often as both parties find necessary to be making progress.
- At least 1 week before your talk: give your talk in front of your mentor who will provide feedback.
- We will expect to have received an electronic copy of your slides day before your presentation.
|Meta-Learning||Johannes Oswald||[pdf] [pdf]|
|GNN: Basic Architectures
GNN: Theoretical Limitations and Power
|GNN: More expressive GNNs||Béni Egressy||[pdf] [pdf]|
|22.03.2022||S Deepak Narayanan
|GNN: More expressive GNNs
GNN: Explainability of GNNs
|Kenza Amara||[pdf] [pdf]|
|GNN: Graph Generation
|Karolis Martinkus||[pdf] [pdf]|
|DnT: Basic Architectures, Ideas, and Metrics
DnT: Disentanglement in RL
|Benjamin Estermann||[pdf] [pdf]|
|12.04.2022||Nathan Corecco||NLP: Basic Architectures||Ard Kastrati||[pdf]|
|NLP: Advanced Architectures
|NLP: Contrastive Learning
NLP+CV: Multimodal Machine Learing
|Zhao Meng||[pdf] [pdf]|
|AL: Program Representations for Learning
AL: Code Search, Code Explanation, Code Completion
|Peter Belcák||[pdf] [pdf]|
|AL: Computational Power, Computational Limits
Reinforcement Learning and Inverse Reinforcement Learning
|24.05.2022||Frédéric Odermatt||From Single-Purpose to Multi-Task & Multi-Modal||