Seminar schedule
Session 1 (10.09)
- Representation Learning with Contrastive Predictive Coding
presenter: Sebastian Szyller
opponent: Khamal Dhakal - Large scale adversarial representation learning
presenter: Paavo Reinikka
opponent: Peter Lövei
Session 2 (17.09)
- A Simple Framework for Contrastive Learning of Visual Representations and Big Self-Supervised Models are Strong Semi-Supervised Learners
presenter: Oliver Struckmeier
opponent: Oliver Järnefelt
- Generative Pretraining from Pixels
presenter: Aleksi Maunu
opponent: Yaroslav Getman
Session 3 (24.09)
- Reformer: The Efficient Transformer
presenter: Sam Splisbury
opponent: Onur Poyraz
- Improving the Gating Mechanism of Recurrent Neural Networks
presenter: Jesse Miettinen
opponent: Andrey Sukhobok
Session 4 (1.10)
- Network Deconvolution
presenter: Markus Ylisiurunen
opponent: Katsiaryna Haitsiukevich - NVAE: A Deep Hierarchical Variational Autoencoder
presenter: Trung Trinh
opponent: Sebastian Szyller
Session 5 (8.10)
- Progressive learning and disentanglement of hierarchical representations
presenter: Pashupati Hedge
opponent: Paavo Reinikka - Weakly-Supervised Disentanglement Without Compromises
presenter: Alex Hämäläinen
opponent: Anirudh Jain
Session 6 (15.10)
- Improving Generative Imagination in Object-Centric World Models
presenter: Maxim Afteniy
opponent: Alex Hämäläinen - Causal Discovery in Physical Systems from Videos
presenter: Sanna Hyvönen
opponent: Trung Trinh
Session 7 (29.10)
- Discovering Symbolic Models from Deep Learning with Inductive Biases
presenter: Anirudh Jain
opponent: Sam Splibury - Learning to Simulate Complex Physics with Graph Networks
presenter: Onur Poyraz
opponent: Mohsen Amidzade
Session 8 (5.11)
- Text as Neural Operator: Image Manipulation by Text Instruction
presenter: Andrey Sukhobok
opponent: Aleksi Maunu - vq-wav2vec: Self-Supervised Learning of Discrete Speech Representations
presenter: Yaroslav Getman
opponent: Sebastian de Peuter
Session 9 (12.11)
- Deep Learning for Symbolic Mathematics
presenter: Daniel Hopkins
opponent: Pashupati Hedge - Neural Arithmetic Units
presenter: Oliver Järnefelt
opponent: Sanna Hyvönen
Session 10 (19.11)
- Concept Learners for Generalizable Few-Shot Learning
presenter: Pius Ladendurber
opponent: Daniel Hopkins - CURL: Contrastive Unsupervised Representations for Reinforcement Learning
presenter: Katsiaryna Haitsiukevich
opponent: Maxim Afteniy
Session 11 (26.11)
- What Can Learned Intrinsic Rewards Capture?
presenter: Peter Lövei
opponent: Pius Ladenburger - A distributional view on multi-objective policy optimization
presenter: Mohsen Amidzade
opponent: Oliver Struckmeier
Session 12 (3.12)
- Ready Policy One: World Building Through Active Learning
presenter: Sebastiaan de Peuter
opponent: Jesse Miettinen - Empirically Verifying Hypotheses Using Reinforcement Learning
presenter: Kamal Dhakal
opponent: Markus Ylisiurunen