Program

Day 1:

Introduction to Deep Learning with Razvan Pascanu

Computer Vision with Natalia Neverova

Tutorial on Supervised Learning with Carl Doersch and Viorica Patraucean (github repo)

Day 2:

Fundamentals of Reinforcement Learning with Diana Borsa

Deep Reinforcement Learning with Doina Precup

Tutorial on Reinforcement Learning with Feryal Behbahani and Gheorghe Comanici (github repo)

Day 3:

Overview of Unsupervised Learning with Mihaela Rosca

Unsupervised learning: kernels, flows and applications with Jacek Tabor

Tutorial on Unsupervised Learning with David Szepesvari and Stanislaw Jastrzebski (github repo)

Day 4:

Using Neural Networks for Modeling and Representing Natural Languages with Tomas Mikolov

QAGS: Question Answering and Generation for evaluating Summarization with Kyunghyun Cho

The Efficient Transformer with Lukasz Kaiser

Tutorials Q&A part1, part2

Day 5:

Discriminative and generative model insights into biological vision with Gergo Orban

Applying Graph Neural Networks at the Bleeding Edge with Petar Velickovic

From classical statistics to modern ML: the lessons of deep learning for generalization and optimization with Mikhail Belkin

Panel on best practices in research with Doina Precup, Nando de Freitas, Yann Dauphin

DeepMind keynote with Irina Higgins

Day 6:

Attention and Memory with Alex Graves

Fireside chat - Challenges in Robotics with Raia Hadsell, moderated by Piotr Mirowski

Fireside chat - Challenges in Reinforcement Learning with David Silver, moderated by Doina Precup

Allegro keynote with Przemyslaw Pobrotyn

Day 7:

Fireside chat with Yoshua Bengio, moderated by Shakir Mohamed

Presentations from ML research groups in Poland and ML associations:

Jagiellonian University Krakow

Wroclaw University

Warsaw University

Poznan University of Technology

Warsaw University of Technology

MLinPL Association

Please find below details about the virtual format of this year's edition. The program features exciting new elements (fireside chats, focused lectures, best practices discussions, mentoring sessions etc.) to facilitate interaction and networking.

The activities will be scheduled as a single track, with the exception of poster sessions where multi-track channels will be used to accommodate a larger number of presentations.

Schedule at a glance (in CEST time zone - Warsaw, Poland):

Lectures: We will use a live Youtube channel and a text-based platform for questions and discussions (rocket chat, sli.do). The session involves the speaker and a moderator. Participants can interact (ask questions or upvote existing questions) through the text-based platform. Length: 1h10 including questions.

Poster sessions: Short pre-recorded video presentations of the posters created by participants will be available to all the participants at the school. Feedback for the posters will be given by text (rocket chat).

Best practices discussions: Panel discussion with renowned researchers in the field, broadcasted on Youtube channel, with questions related to best practices on reading / writing / reviewing papers, running experiments for ML papers etc.

1:1 mentoring sessions (on demand): Participants can subscribe to have a 1:1 mentoring session. There will be a limited number of slots.

Industry keynotes: Format similar to lectures. The session involves the speaker and a moderator, and participants ask questions by text. We will use live Youtube channel with private text channel for questions.

Socialising events: a number of events such as participants introduction sessions, board games, and more will be included in the schedule to facilitate interaction and networking.