CS-E407502 - Special Course in Machine Learning, Data Science and Artificial Intelligence D: Seminar on Deep Learning: 2021, Lectures, 4.11.2021-17.2.2022

Topic outline

• Seminar schedule
  Session 1 (4.11) Reinforcement learning
  

  • Embodied Intelligence via Learning and Evolution
    presenter: Nicola Dainese
    opponent: Aleksi Ikkala
  • The Sensory Neuron as a Transformer: Permutation-Invariant Neural Networks for Reinforcement Learning
    presenter: Yi Zhao
    opponent: Mustafa Mert Celikok

  Session 2 (11.11) Generative models
  

  • Score-Based Generative Modeling through Stochastic Differential Equations
    presenter: Anirudh Jain
    opponent: Maxim Smirnov
  • Diffusion Models Beat GANs on Image Synthesis
    presenter: Severi Rissanen
    opponent: Anirudh Jain

  Session 3 (18.11) Vision
  

  • Image is worth 16x16 words: Transformers for image recognition at scale and MLP-Mixer: An all-MLP Architecture for Vision
    presenter: Mark Seliaev
    opponent: Asael Alonzo
  • Patches Are All You Need?
    presenter: Maria Huovila
    opponent: Kalle Kujanpaa

  Session 4 (25.11) Generative models

  • On Incorporating Inductive Biases into VAEs
    presenter: Siddharth Ramchandran
    opponent: Joel Honkamaa
  • Alias-Free Generative Adversarial Networks (StyleGAN3)
    presenter: Hanna Pankka
    opponent: Maria Huovila

  Session 5 (2.12) Reinforcement learning

  • Transformers are Meta-Reinforcement Learners
    presenter: Hossein Firooz
    opponent: Oliver Jarnefelt
  • Is Curiosity All You Need? On the Utility of Emergent Behaviours from Curious Exploration
    presenter: Christopher Wolanski
    opponent: Aini Putkonen

  Session 6 (9.12) Representation learning

  • Visual Representation Learning Does Not Generalize Strongly Within the Same Domain
    presenter: Aleksi Maunu
    opponent: Hossein Firooz
  • SketchEmbedNet: Learning Novel Concepts by Imitating Drawings
    presenter: Aini Putkonen
    opponent: Mark Seliaev

  Session 7 (13.01) Deep learning architectures

  • Graph Neural Networks with Learnable Structural and Positional Representations
    presenter: Shibei Zhu
    opponent: Hanna Pankka
  • MC-LSTM: Mass-Conserving LSTM
    presenter: Alec Wright
    opponent: Rongzhi Liu

  Session 8 (20.01) Deep learning architectures

  • Evolving Attention with Residual Convolutions
    presenter: Akshenndra Garg
    opponent: Siddharth Ramchandran
  • Systematic generalisation with group invariant predictions
    presenter: Joel Honkamaa
    opponent: Yi Zhao

  Session 9 (27.01) Audio

  • End-to-end Adversarial Text-to-Speech
    presenter: Eloi Moliner
    opponent: Alec Wright
  • DiffWave: A Versatile Diffusion Model for Audio Synthesis
    presenter: Asael Alonzo
    opponent: Severi Rissanen

  Session 10 (03.02) Reinforcement learning

  • Parrot: Data-Driven Behavioral Priors for Reinforcement Learning
    presenter: Kalle Kujanpaa
    opponent: Shibei Zhu
  • Self-supervised Visual Reinforcement Learning with Object Centric Representations
    presenter: Maxim Smirnov
    opponent: Eloi Moliner

  Session 11 (10.02) Biologically inspired architectures

  • Local plasticity rules can learn deep representations using self-supervised contrastive predictions
    presenter: Aleksi Ikkala
    opponent: Aleksi Maunu
  • Towards Biologically Plausible Convolutional Networks
    presenter: Rongzhi Liu
    opponent: Nicola Dainese

  Session 12 (17.02) Reinforcement learning

  • Neural Algorithmic Reasoners are Implicit Planners
    presenter: Mustafa Mert Celikok
    opponent: Akshenndra Garg
  • A learning agent that acquires social norms from public sanctions in decentralized multi-agent settings
    presenter: Oliver Jarnefelt
    opponent: Christopher Wolanski