PhD in Continual Learning (1.0 FTE)

Continual learning is a key aspect of intelligence. The ability to accumulate knowledge by incrementally learning from one’s experiences is an important skill for any agent, natural or artificial, that operates in a changing world. While humans are good continual learners, many AI algorithms—in particular deep neural networks—are not.

We are looking for a highly motivated PhD candidate to study the topic of continual learning, with the aim of developing a deeper conceptual and theoretical understanding of its computational principles. For this PhD position, the intention is to approach continual learning mainly from a deep learning perspective. Depending on the interests of the selected candidate, there is also the opportunity to be involved in (international) collaborations investigating continual learning in the brain.

What are you going to do?

The project’s aims are to gain insights into the computational principles of continual learning, to use those insights to develop novel approaches for continual learning with deep neural networks, and to provide proof-of-principle demonstrations of the benefits of those approaches.

Examples of topics that could be explored in this PhD project:

• Understanding the stability gap. We recently discovered that state-of-the-art methods for continual learning with deep neural networks consistently suffer from temporary yet substantial forgetting when starting to learn something new. A full understanding of this intriguing phenomenon is still lacking.
• Developing optimization-based continual learning methods. The stability gap highlights that solving the continual learning problem requires the development of novel, optimization-based continual learning methods.
• Representational drift. Most existing continual learning methods try to prevent changes to previously learned representations. However, recent neuroscience studies show that, even in the presence of stable behaviour, the underlying neural activity patterns often substantially change over time. Could such representational drift have an important role for continual learning?
• Generalization and continual learning. It appears that continual learning methods successful in mitigating forgetting often impair generalization performance, but a full understanding of the interplay between generalization and continual learning is still missing.

The goal is to contribute fundamental insights to the field of continual learning by producing high-impact research output suitable for publication in top-tier machine learning venues.

Employed PhD candidates are expected to spend 10% of their working hours on teaching and/or supervising students.

You are an enthusiastic and curious researcher with:

• A Master’s degree (completed or near completion) in Computer Science, Artificial Intelligence, Data Science, Computational Neuroscience or another relevant field.
• A solid foundation in machine learning and interest in working on continual learning.
• Strong programming skills, preferably in Python, and familiarity with modern deep learning tools.
• Good analytical and problem-solving abilities, and a critical mindset.
• Very good written and spoken English, as required for scientific communication.
• Motivation to perform high-quality science and publish in leading machine learning venues (e.g., NeurIPS, ICML, ICLR, Nature MI, IEEE TPAMI).
• Evidence of well-executed past research projects (e.g., Master thesis, research assistant position).
• Ability to work both independently and collaboratively in an international research environment.

• 232 vacation hours per year, based on a 38-hour workweek (1.0 FTE). You can also work more or fewer hours in exchange for more or fewer free hours. For example, with a 40-hour workweek, you save 96 extra free hours, and with a 36-hour workweek, you lose 96 hours.
• End-of-year bonus of 8.3% and 8% holiday allowance.
• Extensive opportunities for personal and professional development.

The Faculty of Science and Engineering (FSE) provides teaching and research across a wide range of disciplines, from physics and biology to artificial intelligence, mechanical engineering, and pharmacy. In close collaboration with partners from industry, healthcare, and society, we contribute to the urgent challenges of our time, such as energy, sustainability, digitization, and medical technology. Our community is open and informal, with more than 7,000 students, 1,000 PhD students, and 1,400 staff members from all over the world. If you would like to learn more about the Faculty of Science and Engineering, visit rug.nl/fse. 

This 4-year PhD position is offered at the Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence (https://www.rug.nl/research/bernoulli/). The Bernoulli Institute is a vibrant community with an international outlook, which fosters talent in all its research areas and disciplines and is active in pure and applied science, and (multi)disciplinary research and teaching. Within the Bernoulli Institute, the selected candidate will become a member of the Machine Learning Group, part of the Artificial Intelligence Department, and will work under the supervision of Dr Gido van de Ven.

Do you have any questions or need more information?

Questions about the content of the job?
Gido van de Ven (Assistant Professor): g.m.van.de.ven@rug.nl

Questions about your application process?
Femke Postma (Human Resources Adviser): femke.postma@rug.nl