YAG Pub Lecture Series poster featuring two drinks (a cocktail with lemon slice and a lemonade with lemon slices) on a blue and beige background. Event details include talks by Gido van de Ven on "Forgetting in brains and machines" and Jamal Williams on "Music-evoked memory reactivation and its effects on later retrieval." Event takes place at Vin Natuur on March 4, 2026, with arrival at 16:30, followed by two 15-minute talks with Q&A, and drinks at 18:00. Hosted by University of Groningen and Young Academy Groningen.

If you are in Groningen, don’t miss tomorrow’s #YAGPubLectures! Focus is on the brain, with Gido van de Ven on "Forgetting in brains and machines" and Jamal Williams on "Music-evoked memory reactivation and its effects on later retrieval."

As usual we will start at 17.00 at Vin Natuur!

🚨 We’re hiring! 🚨
Together with Marcus Jeschke and Emilie Mace we are looking for a PhD student to join us for developing AI tools for optogenetic sensory restauration.
Apply now: sinzlab.org/positions/20...
#PhDposition #AI #Neuroprosthetics #ML #NeuroAI #Hiring

It has been claimed that for the performance it doesn’t really matter how the Fisher is computed. But while this holds to some extent for Split MNIST, already with Split CIFAR-10 significant differences in performance emerge.

At ICLR? Come and hear more at poster #483 on Saturday-morning!

On the Computation of the Fisher Information in Continual Learning One of the most popular methods for continual learning with deep neural networks is Elastic Weight Consolidation (EWC), which involves computing the Fisher Information. The exact way in which the Fish...

How do you compute the Fisher when using EWC?

Different ways can be found in the continual learning literature, with the most-used one making rather crude approximations.

This has bothered me (and others!) for a long time, and I finally take this on in an ICLR blogpost: arxiv.org/abs/2502.11756

Why has continual ML not had its breakthrough yet?

In our new collaborative paper w/ many amazing authors, we argue that “Continual Learning Should Move Beyond Incremental Classification”!

We highlight 5 examples to show where CL algos can fail & pinpoint 3 key challenges

arxiv.org/abs/2502.11927

Try out the class-incremental learning with repetition benchmarks of the challenge yourself! github.com/ContinualAI/...

Results without repetition

Results with repetition

These winning strategies clearly outperform experience replay on data streams *with* repetition, but on a “standard” task-based continual learning stream *without* repetition, experience replay performs better.

Schematic of DWGRNet

Schematic of Horde

Schematic of HAT-CIR

A striking outcome of the challenge was that all winning teams used some kind of ensemble-based approach, in which separate sub-networks per task/experience are learned and later combined for making predictions.

Continual learning in the presence of repetition Continual learning (CL) provides a framework for training models in ever-evolving environments. Although re-occurrence of previously seen objects or t…

Does continual learning change when there is repetition in the data stream?

The report of the #CVPR2023 CLVision challenge on **Continual learning in the presence of repetition** is out in Neural Networks. #OpenAccess

www.sciencedirect.com/science/arti...

Aligning Generalisation Between Humans and Machines Recent advances in AI -- including generative approaches -- have resulted in technology that can support humans in scientific discovery and decision support but may also disrupt democracies and target...

This is one of the most important arguments in the AI discourse, I think: With a large group of experts we explain why generalization of humans and machines works very differently. This is a fundamental points that has crucial implications for language models, as well: arxiv.org/abs/2411.15626

Is generalisation a process, an operation, or a product? 🤨

Read about the different ways generalisation is defined, parallels between humans & machines, methods & evaluation in our new paper: arxiv.org/abs/2411.15626

co-authored with many smart minds as a product of Dagstuhl 🙏🎉

Thanks Simon! I’d be keen to be added as well ☺️