# Talks

**NB:** Click title to toggle abstract.

^{*} These sessions will be recorded. We will upload the recorded talks to the LMS youtube channel after the end of this symposium.

Time (BST) | Event |
---|---|

8:50 - 9:00 | Opening remarks by Johannes Zimmer Video |

9:00 - 9:45 | ## Stephane Chretien |

9:45 - 10:30 | ## Johannes Schmidt-Hieber |

Break | |

13:00 - 13:45 | ## Lénaïc Chizat |

13:45 - 14:30 | ## Varun Kanade: Implicit Regularization Properties Early-Stopped Gradient-based Algorithms. Slides |

Break | |

15:30 - 16:15 | ## Weinan E |

16:15 - 17:00 | ## Brynjulf Owren |

Time (BST) | Event |
---|---|

9:00 - 9:45 | ## Coralia Cartis: Dimensionality reduction techniques for large-scale optimization problems Video |

9:45 - 10:30 | ## Marcelo Pereyra |

Break | |

13:00 - 13:45 | ## Elena Celledoni: Deep learning as optimal control and structure preserving deep learning Slides |

13:45 - 14:30 | ## Pierfrancesco Urbani: Tracking the dynamics of gradient based algorithms in high dimensional optimization. |

Break | |

15:30 - 16:15 | ## Spencer Thomas |

16:15 - 17:45 | Poster session: Posters 15-26 |

Time (BST) | Event |
---|---|

9:00 - 9:45 | ## Jong Chul Ye |

9:45 - 10:30 | ## Jonas Adler: A case for provable properties in neural networks Slides |

Break | |

13:00 - 13:45 | ## Philipp Petersen |

13:45 - 14:30 | ## Simon Arridge |

Break | |

15:30 - 16:15 | ## Alhussein Fawzi: Proving inequalities with deep learning. Slides Video |

16:15 - 17:00 | ## Carola-Bibiane Schönlieb |

Time (BST) | Event |
---|---|

9:00 - 10:30 | Poster session: Posters 1-14 |

10:30 - 11:15 | ## Ivan Tyukin |

Break | |

13:00 - 13:45 | ## Anders Hansen |

13:45 - 14:30 | ## Andrew Fitzgibbon |

Break | |

15:30 - 16:15 | ## Yuejie Chi |

16:15 - 17:00 | ## Nadia Drenska |

Break | |

18:00 - 19:00 | ## Carola-Bibiane Schönlieb (Public lecture) |

Time (BST) | Event |
---|---|

9:00 - 9:45 | ## Ben Leimkuhler |

9:45 - 10:30 | ## Martin Benning |

Break | |

13:00 - 13:45 | ## Erik Bekkers |

13:45 - 14:30 | ## Sofia Olhede |

Break | |

15:30 - 16:15 | ## Aretha Teckentrup |

16:15 - 17:00 | ## Matthew Thorpe |

# Posters

The poster session will be held on Gather.Town (a link has been sent out to registered participants). Posters 15-26 will be presented on Tuesday and Posters 1-14 on Thursday, but feel free to enter the poster room anytime to browse.

Tips for the poster session:

- Please use Chrome or Firefox.
- Zooming out on your browser will actually make the poster screen larger
- Please mute yourself if you are not speaking.

**Poster Prizes**
We are pleased to announce the following winners of the poster prize:

- First place to
**Allard Hendriksen**for*Noise2Inverse: Deep tomographic denoising without high-quality target data* - Second place to
**Salvatore Danilo Riccio**for*Robustness of Runge-Kutta networks against adversarial attacks* - Third place to
**Eliot Ayache and Tanmoy Laskar**for*Machine learning applications in High-Energy Time-Domain Astrophysics*

ID | Presenter | Title |
---|---|---|

001 | Eliot Ayache / Tanmoy Laskar (University of Bath) | Machine learning applications in High-Energy Time-Domain Astrophysics Poster Video |

002 | Pasquale Cascarano (Università di Bologna) | Deep Plug-and-Play Gradient Method for Super-Resolution Poster Video |

003 | Eric Baruch Gutierrez Castillo (University of Bath) | On Primal-Dual Algorithms for Nonsmooth Large-Scale Machine Learning Poster |

004 | Dongdong Chen (University of Edinburgh) | Deep Plug-and-Play Network for Compressive MRF reconstruction Poster |

005 | ||

006 | Jacob Deasy (University of Cambridge) | Closed-form differential entropy of a multi-layer perceptron variant Poster |

007 | Margaret Duff (University of Bath) | Solving Inverse Imaging Problems with Generative Machine Learning Models Poster Video |

008 | David Fernandes (University of Bath) | Unsupervised Learning with GPs and SDEs Poster |

009 | Allen Hart (University of Bath) | Using Echo State Networks to solve Stochastic Optimal Control Problems Poster |

010 | Allard Hendriksen (Centrum Wiskunde & Informatica (CWI), Amsterdam) | Noise2Inverse: Deep tomographic denoising without high-quality target data Poster |

011 | Johannes Hertrich (TU Berlin) | Parseval Proximal Neural Networks Poster |

012 | Nicolas Keriven (CNRS, GIPSA-lab) | Universal Invariant and Equivariant Graph Neural Networks Poster |

013 | Youngkyu Lee (KAIST) | A parareal neural network emulating parallel-in-time algorithm Poster |

014 | Peter Mathé (Weierstrass Institute) / Abhishake Rastogi (University of Potsdam) | Inverse learning in Hilbert scales Poster |

015 | Janith Petangoda (Imperial College London) | Transfer Learning: An application of Foliations Poster |

016 | Gabriele Incorvaia (University of Manchester) | A deep learning application for Through-the-Wall Radar Imaging Poster |

017 | Ilan Price (University of Oxford) | Trajectory growth through deep random ReLU networks. Poster Video |

018 | Abhishake Rastogi (University of Potsdam) | Regularization schemes for statistical inverse problems Poster |

019 | Salvatore Danilo Riccio (Queen Mary University of London) | Robustness of Runge-Kutta networks against adversarial attacks Poster Video |

020 | Paul Russell (University of Bath) | Learning to solve Rubik’s cube Poster |

021 | Malena Sabaté Landman (University of Bath) | Iteratively Reweighted Flexible Krylov methods for Sparse Reconstruction Poster Video |

022 | Silvester Sabathiel (NTNU Trondheim) | A computational model of learning to count in a multimodal, interactive environment. Poster Video |

023 | Ferdia Sherry (University of Cambridge) | Equivariant neural networks for inverse problems in imaging Poster |

024 | Giuseppe Ughi (University of Oxford) | A Model-Based Derivative-Free Approach to Black-Box Adversarial Examples Poster |

025 | Xiaoyu Wang (University of Cambridge) | Gradient-based training of non-smooth neural networks Poster Video |

026 | Kelvin Shuangjian Zhang (ENS Paris) | Wasserstein control of mirror Langevin Monte Carlo Poster |

# Public Lecture

**Speaker:** Carola-Bibiane Schönlieb, Professor at the University of Cambridge.

**Title:** Looking into the black box: how mathematics can help to turn deep learning inside out

**Abstract:** Deep learning has had a transformative impact on a wide range of tasks related to Artificial Intelligence, ranging from computer vision and speech recognition to playing games. Still, the inner workings of deep neural networks are far from clear, and designing and training them is seen as almost a black art. In this talk we will try to open this black box a little bit by using mathematical structure of neural networks described by so-called differential equations and mathematical optimisation. The talk is furnished with several examples in image analysis and computer vision, ranging from biomedical imaging to remote sensing.

**Time:** 18.00 BST, Thursday 6th August 2020.

Sign up here. A Zoom link will be sent to registered participants on the day of the lecture.