Matthijs Pals

Got prov. approval for 2 major grants in Neuro-AI & Dynamical Systems Reconstruction, on learning & inference in non-stationary environments, out-of-domain generalization, and DS foundation models. To all AI/math/DS enthusiasts: Expect job announcements (PhD/PostDoc) soon! Feel free to get in touch.

Jelmer Borst and I are looking for a PhD candidate to build an EEG-based model of human working memory! This is a really cool project that I've wanted to kick off for a while, and I can't wait to see it happen. Please share and I'm happy to answer any Qs about the project!
www.rug.nl/about-ug/wor...

Null and Noteworthy: Neurons tracking sequences don’t fire in order Instead, neurons encode the position of sequential items in working memory based on when they fire during ongoing brain wave oscillations—a finding that challenges a long-standing theory.

The neurons that encode sequential information into working memory do not fire in that same order during recall, a finding that is at odds with a long-standing theory. Read more in this month’s Null and Noteworthy.

By @ldattaro.bsky.social

#neuroskyence

www.thetransmitter.org/null-and-not...

Abstract rule learning promotes cognitive flexibility in complex environments across species Nature Communications - Whether neurocomputational mechanisms that speed up human learning in changing environments also exist in other species remains unclear. Here, the authors show that both...

How do animals learn new rules? By systematically testing diff. behavioral strategies, guided by selective attn. to rule-relevant cues: rdcu.be/etlRV
Akin to in-context learning in AI, strategy selection depends on the animals' "training set" (prior experience), with similar repr. in rats & humans.

Out today in @nature.com: we show that individual neurons have diverse tuning to a decision variable computed by the entire population, revealing a unifying geometric principle for the encoding of sensory and dynamic cognitive variables.
www.nature.com/articles/s41...

Our new preprint 👀

Memory by a thousand rules: Automated discovery of functional multi-type plasticity rules reveals variety & degeneracy at the heart of learning Synaptic plasticity is the basis of learning and memory, but the link between synaptic changes and neural function remains elusive. Here, we used automated search algorithms to obtain thousands of str...

We just pushed “Memory by a 1000 rules” onto bioRxiv, where we use clever #ML to find #plasticity quadruplets (EE, EI, IE, II) that learn basic stability in spiking nets. Why is it cool? We find 1000s!! of solutions, and they don’t just stabilise. They #memorise! www.biorxiv.org/content/10.1...

A wide shot of approximately 30 individuals standing in a line, posing for a group photograph outdoors. The background shows a clear blue sky, trees, and a distant cityscape or hills.

Great news! Our March SBI hackathon in Tübingen was a huge success, with 40+ participants (30 onsite!). Expect significant updates soon: awesome new features & a revamped documentation you'll love! Huge thanks to our amazing SBI community! Release details coming soon. 🥁 🎉

Please RT🙏

Reach out if you want to help understand cognition by modelling, analyzing and/or collect large scale intracortical data from 👩🐒🐁

We're a friendly, diverse group (n>25) w/ this terrace 😎 in the center of Paris! See👇 for + info about the lab

We have funding to support your application!

🎓Hiring now! 🧠 Join us at the exciting intersection of ML and Neuroscience! #AI4science
We’re looking for PhDs, Postdocs and Scientific Programmers that want to use deep learning to build, optimize and study mechanistic models of neural computations. Full details: www.mackelab.org/jobs/ 1/5

Re-posting is appreciated: We have a fully funded PhD position in CMC lab @cmc-lab.bsky.social (at @tudresden_de). You can use forms.gle/qiAv5NZ871kv... to send your application and find more information. Deadline is April 30. Find more about CMC lab: cmclab.org and email me if you have questions.

Compositional simulation-based inference for time series Amortized simulation-based inference (SBI) methods train neural networks on simulated data to perform Bayesian inference. While this strategy avoids the need for tractable likelihoods, it often requir...

Excited to present our work on compositional SBI for time series at #ICLR2025 tomorrow!

If you're interested in simulation-based inference for time series, come chat with Manuel Gloeckler or Shoji Toyota

at Poster #420, Saturday 10:00–12:00 in Hall 3.

📰: arxiv.org/abs/2411.02728

Excited to announce that our paper on "Comparing noisy neural population dynamics using optimal transport distances" has been selected for an oral presentation in #ICLR2025 (1.8% top papers). Check the thread for paper details (0/n).

Presentation info: iclr.cc/virtual/2025....

Happening tomorrow morning :).

The @mackelab.bsky.social is represented at @cosynemeeting.bsky.social #cosyne2025 in Montreal

with 3 posters, 2 workshop talks, and a main conference contributed talk (for the very first time in Mackelab history 🎉)!

Out now 'in print'--- a true labor of love, in more ways than one. See the paper and press-release below!

Also, go and see @matthijspals.bsky.social's talk at #Cosyne, where he will talk about related/follow up work!

uni-tuebingen.de/en/research/...

Our study on sequence working memory using human spiking data and RNNs, is finally published :). Check it out! 👇

In the physical world, almost all information is transmitted through traveling waves -- why should it be any different in your neural network?

Super excited to share recent work with the brilliant @mozesjacobs.bsky.social: "Traveling Waves Integrate Spatial Information Through Time"

1/14

Together with @dendritesgr.bsky.social, we’ll be hosting a tutorial on constructing and optimizing biophysical models (via Jaxley & DendroTweaks) 🚀

Join us in Florence if you like dendrites, biophysics, or optimization!

1) Some exciting science in turbulent times:

How do mice distinguish self-generated vs. object-generated looming stimuli? Our new study combines VR and neural recordings from superior colliculus (SC) 🧠🐭 to explore this question.

Check out our preprint doi.org/10.1101/2024... 🧵

🚨Excited to share my first @biorxivpreprint.bsky.social🚨 with the amazing Smith lab, @jbarbosa.org and Albert Compte who made this work possible.

We show that 🐒prefrontal hemispheres combine redundancy (for precision) & weak connections (for capacity) for supporting spatial working memory (WM). 1/🧵

Hi, could you add me? :D

Pre-print 🧠🧪
Is mechanism modeling dead in the AI era?

ML models trained to predict neural activity fail to generalize to unseen opto perturbations. But mechanism modeling can solve that.

We say "perturbation testing" is the right way to evaluate mechanisms in data-constrained models

1/8

Talk to @vetterj.bsky.social and @gmoss13.bsky.social about sourcerer at #Neurips2024 today!
📍Poster #4006 (East; 11 am PT)

That Eq. 3 can be low-D, while the observations (Eq. 5) can be very high dimensional is kind of a nice way to reconcile computational models of low-D dynamics with experimental recordings that are high-D, I think @bio-emergent.bsky.social has some nice work on this :).

I agree that when very concretely linking the RNNs to data, this is important! There is a large line of work making the comparison more on the level of low-D "population" dynamics, where the exact biological equivalent of the neuron states is less important.

Mathematical Equivalence of Two Common Forms of Firing-Rate Models of Neural Networks We demonstrate the mathematical equivalence of two commonly used forms of firing-rate model equations for neural networks. In addition, we show that what is commonly interpreted as the firing rate in ...

Unless I’m missing something, you could still see 1 as describing a low-pass filtered version of the spike dynamics, see last paragraph of: pmc.ncbi.nlm.nih.gov/articles/PMC...

We reduce the cost of finding all fixed-points in piece-wise linear low-rank RNNs from 2^N to O(N^R)! This is part of our recent NeurIPS paper, presented tomorrow:
bsky.app/profile/mack...
Special thanks to Manuel Gloeckler for helping write out a proof! 7/7

But not all of the 2^N regions do! Each neuron partitions the R-dim subspace of dynamics with a hyperplane (here a line). N hyperplanes can partition R-dim space into at most O(N^R) regions. We can thus reduce our search space by only solving for fixed points in these!
6/7

Some of the the 2^N regions with linear dynamics intersect the subspace in which the dynamics unfold (span U), as the one here: 5/7