HICT: Doctoral Candidate in Developing Novel Symmetry-Learning Algorithms for out-of-distribution generalization

The Helsinki Doctoral Education Network in Information and Communications Technology (HICT) is a collaborative doctoral education network hosted jointly by Aalto University and the University of Helsinki, the two leading universities within this area in Finland. The network serves as a collaboration platform for doctoral education combining all the relevant subfields of computer science and information technology at Aalto University and the University of Helsinki. It involves at present 80 professors and almost 300 doctoral students, and the participating units graduate altogether more than 40 new doctors each year.

We offer the possibility to join world-class research groups in both universities with multiple interesting research projects to choose from. The list of all open Doctoral Candidate positions in this joint call can be found here: https://hict.fi/admissions/

HICT Doctoral Candidate in Developing Novel Symmetry-Learning Algorithms for out-of-distribution generalization

Supervisor(s): Stephane Deny (Aalto University)

Traditional deep learning methods typically require very large training datasets, and they only generalize well to data coming from the very same distribution as the training set. These two requirements make deep learning methods difficult to apply out-of-the-box to many industry problems where the training data is limited, and where the use-case data characteristics might not be represented in the training set. Recently, alternative deep learning architectures called equivariant neural networks have been proposed to tackle these issues. They consist in encoding prior knowledge of the symmetries of a problem into the architecture of a network. Applied to small medical imaging datasets, these methods are currently state-of-the-art and outperform traditional deep networks. However, these methods have been developed to build invariance to well-known symmetries found in images (e.g., translation, scale, rotation), and are therefore not directly applicable to other types of data, for which the specific symmetries of the problems are unknown (e.g., biomedical data). In this project, we will build upon my recent work (Bouchacourt et al., 2021: https://ai.facebook.com/blog/building-ai-that-can-understand-variation-in-the-world-around-us ) to develop novel algorithms able to learn the symmetries of a problem from the data itself. The expected outcome is to learn data-efficient representations that can generalize to data characteristics never seen during training.

About me:  I am a new assistant professor in the Department of Neuroscience and Biomedical Engineering and the Department of Computer Science at Aalto University. During my PhD (Vision Institute of Paris) and postdocs (Stanford, Meta AI), I have worked on retinal interfaces for blind patients, neural data analysis and state-of-the-art methods for self-supervised learning and symmetry-learning (more: https://sites.google.com/view/stephanedeny/home )

Preferred skills: Some experience with a deep learning language such as PyTorch or Tensorflow. Some interest in the topic.

Ready to apply?

If you want to join our community and start your journey towards Doctoral dissertation submit your application via our electronic recruitment system. The application form will close 6th February 2022 at 23:59 Finnish time (UTC +2).

If you are currently working at Aalto University, please submit your application via HR System Workday (Career – Find jobs – Apply).

Please be prepared to add the following attachments:

• Cover letter
• CV
• Transcript of your Master’s studies

You will be asked to provide also the contact details of 2-3 academic referees.

What we offer?

HICT is a network of world-class professors and researchers with huge passion for their field of study. In HICT you will become part of enthusiastic and professional team of bold and innovative thinkers.

During your Doctoral studies you will be a full-time employee in either Aalto University or the University of Helsinki depending on the supervisor and the research area. You will be covered by occupational health care based on the employment contract.

More information:

General information about the joint HICT call and HICT: https://hict.fi/

General questions about HICT: Christina Sirviö, HICT team

General questions about recruitment process: Sanni Kirmanen, Aalto University HR

[email protected]

More about Aalto University:

Aalto.fi
twitter.com/aaltouniversity
facebook.com/aaltouniversity
instagram.com/aaltouniversity