ELLIS PhD & PostDoc Program

Thumb ticker md steffen schneider

Adaptation and Robustness in Brains and Machines

Steffen Schneider (Ph.D. Student)

Understanding the mechanisms underlying robust learning and efficient adaptation is an open problem both in neuroscience and machine learning. While robustness and domain adaptation in ML is commonly studied with computer vision tasks, adaptation research in neuroscience has been traditionally carried out in sensori...

Primary Host: Matthias Bethge (University of Tübingen, Germany)
Exchange Host: Mackenzie Mathis (The Rowland Institute at Harvard, Cambridge, MA, USA & EPFL, Switzerland)
PhD Duration: 01 November 2019 - 31 October 2022
Exchange Duration: 18 February 2020 - 31 March 2020
Thumb ticker md portrait

Flexible Densities for Deep Generative Models

Didrik Nielsen (Ph.D. Student)

Probability distributions play a central role in machine learning. For probabilistic modeling, they are used as likelihoods and prior distributions, whereas in variational inference, they are employed as approximate posterior distributions. The probability distributions typically used in practice tend to be simple...

Primary Host: Ole Winther (University of Copenhagen and Technical University of Denmark)
Exchange Host: Max Welling (University of Amsterdam & Qualcomm)
PhD Duration: 01 January 2019 - 31 December 2019
Exchange Duration: 13 January 2020 - 29 May 2020
Thumb ticker md giambattista parascandolo  phd student

Generalization out of distribution

Giambattista Parascandolo (Ph.D. Student)

While current techniques in machine learning have been showing tremendous success at generalization in the i.i.d. setting when large quantities of data and compute are available, performance consistently drops as soon as we try to extend our models to data out of distribution. Tasks such as transfer learning, meta-l...

Primary Host: Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
Exchange Host: Thomas Hofmann (ETH Zürich)
PhD Duration: 01 March 2017 - 01 March 2021
Exchange Duration: 01 February 2020 - 01 February 2021
Thumb ticker md brandstetter johannes

Geometric deep learning and partial differential equations

Johannes Brandstetter (PostDoc)

Partial differential equations (PDEs) are used in physics, engineering and many other scientific disciplines. While their numerical solutions have been a longstanding challenge, deep learning methods offer an appealing meshfree approach. On the other hand, PDEs are a language in which we can express inductive biases...

Primary Host: Sepp Hochreiter (JKU Linz)
Exchange Host: Max Welling (University of Amsterdam & Qualcomm)
PostDoc Duration: 01 July 2018 - 30 June 2024
Exchange Duration: 01 September 2020 - 31 August 2022
Thumb ticker md gresele luigi

Independent Component Analysis: linear and nonlinear, single and multi-view. Identifiability and estimation algorithms

Luigi Gresele (Ph.D. Student)

Independent Component Analysis (ICA) provides a principled framework for unsupervised feature extraction and blind source separation, with ubiquitous applications in signal processing, astronomy and neuroimaging. In the multi-view setting, the aim is to extract common sources of variability from multiple related obs...

Primary Host: Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
Exchange Host: Aapo Hyvärinen (University of Helsinki and University College London)
PhD Duration: 01 October 2017 - Ongoing
Exchange Duration: 01 October 2019 - 31 December 2019
Thumb ticker md despoina paschalidou

Learning Deep Models with Primitive-based Representations

Despoina Paschalidou (Ph.D. Student)

My research so far seeks to give an answer to a very simple question, how can we teach machines to learn to see in 3D? Or in other words, what is the best representation, that would allow us to capture the world such that a machine would be able to robustly perceive it? Humans develop a common-sense understanding of...

Primary Host: Andreas Geiger (University of Tübingen, Germany, Max Planck Institute for Intelligent Systems)
Exchange Host: Luc Van Gool (ETH Zürich, KU Leuven)
PhD Duration: 01 April 2017 - 31 March 2021
Exchange Duration: 01 February 2019 - 31 July 2020
Thumb ticker md w01 small

Learning Robotic Manipulation from Instructional Videos

Vladimir Petrik (PostDoc)

The objective of the project is to enable robots to learn new manipulation skills from instructional videos available online. We will study how instructional videos could be used to overcome the sparse reward problem in reinforcement learning. The sparse reward complicates reinforcement learning by assigning the rew...

Primary Host: Josef Sivic (INRIA and Czech Technical University)
Exchange Host: Ivan Laptev (INRIA, FR)
PostDoc Duration: 01 April 2019 - 31 March 2020
Exchange Duration: 01 October 2019 - 31 January 2020
Thumb ticker md john bradshaw  phd student

Machine Learning for Molecules

John Bradshaw (Ph.D. Student)

Machine Learning has enormous potential in augmenting scientists' capabilities in discovering novel drugs or new materials. To achieve this we need to develop ML models that can accurately predict properties of molecules and their interactions, as well as techniques that enable intelligent searching of discrete and ...

Primary Host: Jose Miguel Hernandez Lobato (University of Cambridge)
Exchange Host: Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
PhD Duration: 01 October 2016 - Ongoing
Exchange Duration: 01 November 2018 - 31 December 2019
Thumb ticker md niki kilbertus  phd student

Socially Beneficial Machine Learning

Niki Kilbertus (Ph.D. Student)

As machine learning touches upon all areas of our daily lives, it is increasingly deployed to make or support consequential decisions about individuals. Such applications raise concerns about privacy violations, the fairness of algorithms, as well as the long-term impact automated decisions might have on individuals...

Primary Host: Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
Exchange Host: Carl Rasmussen (University of Cambridge)
PhD Duration: 01 October 2016 - Ongoing
Exchange Duration: 01 September 2017 - 30 June 2018
Thumb ticker md anant raj

Stochastic Convex Optimization for Over-Parametrized models

Anant Raj (Ph.D. Student)

Over-parametrized models are frequently occurring phenomena in machine learning which comes with nice properties. In this work, we investigate methods to optimize such models. Our goal is to show faster convergence rate for traditional 1st order methods on such problems without extra assumptions and with similar com...

Primary Host: Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
Exchange Host: Francis Bach (INRIA, Ecole Normale Superieur, Paris, France)
PhD Duration: 01 August 2015 - Ongoing
Exchange Duration: 04 October 2019 - 31 March 2020
Thumb ticker md paul kishan rubenstein  phd student

Theory of latent feature learning

Paul Kishan Rubenstein (Ph.D. Student)

Low dimensional, abstract or otherwise 'simple' structure occurs widely across machine learning. This project advances theoretical understanding in a variety of areas. These are: causality, in which a theory of micro-macro abstractions is developed; independent component analysis, in which new identifiability result...

Primary Host: Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
Exchange Host: Carl Rasmussen (University of Cambridge)
PhD Duration: 01 October 2015 - 30 June 2020
Exchange Duration: 01 October 2015 - 30 September 2016
Thumb ticker md alessandro davide ialongo  phd student

Uncertainty Quantification in Dynamical Systems

Alessandro Davide Ialongo (Ph.D. Student)

Many real-world systems are not static, they evolve through time. Modelling them as dynamical systems enables us to correctly account for non-stationarity and is a natural choice for sequential datasets. Especially in the low data regime, correctly quantifying predictive uncertainty is crucial to ensure we do not ta...

Primary Host: Carl Rasmussen (University of Cambridge)
Exchange Host: Bernhard Schölkopf (Max Planck Institute for Intelligent Systems)
PhD Duration: 01 October 2016 - 31 May 2021
Exchange Duration: 01 November 2018 - 30 April 2020
Thumb ticker md mancini massimiliano

Zero shot adaptation and learning

Massimiliano Mancini (Ph.D. Student)

Domain Adaptation is a transfer learning scenario where the goal is to build a model addressing a task, e.g. classification, in a target domain with no or few images are labeled. Given a large amount of labeled data in a domain, i.e. the source, with a different input distribution from the target, e.g. synthetic to ...

Primary Host: Barbara Caputo (Politecnico di Torino & Italian Institute of Technology)
Exchange Host: Zeynep Akata (University of Tübingen and Max Planck Institute for Informatics)
PhD Duration: 01 November 2016 - 31 October 2020
Exchange Duration: 01 March 2020 - 30 June 2020