Name: Second HSE-Yandex autumn school on generative models
Start: 2020-11-24T09:45:00+03:00
End: 2020-11-27T21:00:00+03:00
Location: NRU HSE, Moscow

Second HSE-Yandex autumn school on generative models

from Tuesday, November 24, 2020 (9:45 AM) to Friday, November 27, 2020 (9:00 PM)

Monday, November 23, 2020
Tuesday, November 24, 2020

9:45 AM Welcome
Welcome
9:45 AM - 10:00 AM
Room: R602

10:00 AM Introduction to machine learning
Introduction to machine learning
10:00 AM - 11:00 AM
Room: R602

11:00 AM Coffee Break
Coffee Break
11:00 AM - 11:30 AM
Room: S812

11:30 AM Introduction to machine learning. Practice
Introduction to machine learning. Practice
11:30 AM - 12:30 PM
Room: R602

12:30 PM Lunch
Lunch
12:30 PM - 2:00 PM

2:00 PM Introduction to neural networks
Introduction to neural networks
2:00 PM - 3:00 PM
Room: R602

3:00 PM Tea Break
Tea Break
3:00 PM - 3:30 PM
Room: S812

3:30 PM Introduction to neural networks. Practice
Introduction to neural networks. Practice
3:30 PM - 4:30 PM
Room: R602
4:30 PM Unusual Effects in Modern Deep Neural Networks - Dmitriy Vetrov (HSE)
Unusual Effects in Modern Deep Neural Networks
- Dmitriy Vetrov (HSE)
4:30 PM - 5:30 PM
Room: R602
Wednesday, November 25, 2020

10:00 AM Introduction, the metrics for the generative models
Introduction, the metrics for the generative models
10:00 AM - 11:00 AM
Room: R608

11:00 AM Coffee Break
Coffee Break
11:00 AM - 11:30 AM
Room: S812
11:30 AM Generative adversarial networks - Artem Maevskiy (HSE University)
Generative adversarial networks
- Artem Maevskiy (HSE University)
11:30 AM - 12:30 PM
Room: R608
12:30 PM Group Photo
Group Photo
12:30 PM - 12:40 PM
Room: R602

12:40 PM Lunch
Lunch
12:40 PM - 2:00 PM
2:00 PM Generative Adversarial Networks. Practice
Generative Adversarial Networks. Practice
2:00 PM - 3:00 PM
Room: R608
Contributions

2:00 PM Generative adversarial networks - Artem Maevskiy (HSE University)
3:00 PM Tea Break
Tea Break
3:00 PM - 3:30 PM
Room: S812

3:30 PM Introduction to the normalizing flows. Lecture
Introduction to the normalizing flows. Lecture
3:30 PM - 4:30 PM
Room: R608

4:30 PM Tea Break
Tea Break
4:30 PM - 4:45 PM
Room: R602

4:45 PM Normalizing flows, practice
Normalizing flows, practice
4:45 PM - 5:45 PM
Room: R608
Thursday, November 26, 2020

10:00 AM An introduction to bandit problems
An introduction to bandit problems
10:00 AM - 11:00 AM
Room: R602

11:00 AM Coffee Break
Coffee Break
11:00 AM - 11:30 AM
Room: S812

11:30 AM An introduction to bandit problems
An introduction to bandit problems
11:30 AM - 12:30 PM
Room: R602

12:30 PM Lunch
Lunch
12:30 PM - 2:00 PM

2:00 PM An introduction to bandit problems
An introduction to bandit problems
2:00 PM - 3:00 PM
Room: R602

3:00 PM Tea Break
Tea Break
3:00 PM - 3:30 PM
Room: S812

3:30 PM An introduction to bandit problems
An introduction to bandit problems
3:30 PM - 4:30 PM
Room: R602
4:30 PM Robust mean estimation - Nikita Zhivotovskiy (Google, Higher School of Economics)
Robust mean estimation
- Nikita Zhivotovskiy (Google, Higher School of Economics)
4:30 PM - 5:30 PM
Room: R602
Friday, November 27, 2020
10:00 AM Synthetic data and generative models in road sign recognition and other computer vision problems - Anton Konushin (Higher School of Economics)
Synthetic data and generative models in road sign recognition and other computer vision problems
- Anton Konushin (Higher School of Economics)
10:00 AM - 10:40 AM
Room: R602
10:40 AM Latent-Space Laplacian Pyramids for Adversarial Representation Learning with 3D Point Clouds - Alexey Artemov (Yandex)
Latent-Space Laplacian Pyramids for Adversarial Representation Learning with 3D Point Clouds
- Alexey Artemov (Yandex)
10:40 AM - 11:20 AM
Room: R602 Constructing high-quality generative models for 3D shapes is a fundamental task in computer vision with diverse applications in geometry processing, engineering, and design. Despite the recent progress in deep generative modelling, synthesis of finely detailed 3D surfaces, such as high-resolution point clouds, from scratch has not been achieved with existing approaches. In this work, we propose to employ the latent-space Laplacian pyramid representation within a hierarchical generative model for 3D point clouds. We combine the recently proposed latent-space GAN and Laplacian GAN architectures to form a multi-scale model capable of generating 3D point clouds at increasing levels of detail. Our evaluation demonstrates that our model outperforms the existing generative models for 3D point clouds.
11:20 AM Coffee Break
Coffee Break
11:20 AM - 11:40 AM
Room: S813
11:40 AM Deep Generative Models for Knowledge Transfer - Evgeny Burnaev (Skoltech)
Deep Generative Models for Knowledge Transfer
- Evgeny Burnaev (Skoltech)
11:40 AM - 12:20 PM
Room: R602
12:20 PM Optimal transport and its application in bioinformatics - Alexandra Suvorikova (Higher School of Economics)
Optimal transport and its application in bioinformatics
- Alexandra Suvorikova (Higher School of Economics)
12:20 PM - 1:00 PM
Room: R602
1:00 PM Lunch
Lunch
1:00 PM - 2:10 PM
2:10 PM How to reveal hidden geometric structures in data - Evgeny Stepanov (Higher School of Economics)
How to reveal hidden geometric structures in data
- Evgeny Stepanov (Higher School of Economics)
2:10 PM - 2:50 PM
Room: R602
2:50 PM Variational Inference: Mean Field, Normalizing Flows and beyond - Maxim Panov (Skoltech, HSE)
Variational Inference: Mean Field, Normalizing Flows and beyond
- Maxim Panov (Skoltech, HSE)
2:50 PM - 3:30 PM
Room: R602 In this talk we are going to discuss Variational approach to Bayesian Inference. We will start by reviewing standard mean-field approximations to the posterior as well as more powerful methods such as normalizing flows. Then we will discuss a new computationally efficient method to combine Variational Inference (VI) with Markov Chain Monte Carlo (MCMC). This approach can be used with generic MCMC kernels, but is especially well suited to MetFlow, a novel family of MCMC algorithms we introduce, in which proposals are obtained using Normalizing Flows. The marginal distribution produced by such MCMC algorithms is a mixture of flow-based distributions, thus drastically increasing the expressivity of the variational family. Unlike previous methods following this direction, our approach is amenable to the reparametrization trick and does not rely on computationally expensive reverse kernels
3:30 PM Big GANs Are Watching You: Towards Unsupervised Object Segmentation with Off-the-Shelf Generative Models - Stanislav Morozov (Yandex)
Big GANs Are Watching You: Towards Unsupervised Object Segmentation with Off-the-Shelf Generative Models
- Stanislav Morozov (Yandex)
3:30 PM - 4:10 PM
Room: R602 Since collecting pixel-level groundtruth data is expensive, unsupervised visual understanding problems are currently an active research topic. In particular, several recent methods based on generative models have achieved promising results for object segmentation and saliency detection. However, since generative models are known to be unstable and sensitive to hyperparameters, the training of these methods can be challenging and time-consuming. In this work, we introduce an alternative, much simpler way to exploit generative models for unsupervised object segmentation. First, we explore the latent space of the BigBiGAN -- the state-of-the-art unsupervised GAN, which parameters are publicly available. We demonstrate that object saliency masks for GAN-produced images can be obtained automatically with BigBiGAN. These masks then are used to train a discriminative segmentation model. Being very simple and easy-to-reproduce, our approach provides competitive performance on common benchmarks in the unsupervised scenario.
4:10 PM Tea Break
Tea Break
4:10 PM - 4:30 PM
Room: R602
4:30 PM
4:30 PM - 5:50 PM
Room: R602
Contributions

4:30 PM How to reveal hidden geometric structures in data - Evgeny Stepanov (Higher School of Economics)
5:50 PM The end
The end
5:50 PM - 6:00 PM
Room: R602

7:00 PM Evening session -- Cancelled
Evening session -- Cancelled
7:00 PM - 8:30 PM
Room: R602

Choose timezone