My new lab @USherbrooke is hiring (MSc, PhD)! * Cool research on neuroimaging, epilepsy, brain development, etc. * Located in a brand new building (lots of natural light) * Worldwide collaborations * Fun & inclusive environment * FR/EN Find out more:

I’m proud to say that “Relating sex-bias in human cortical and hippocampal microstructure to sex hormones” is now out in @NatureComms 🥳! https://t.co/pNYC2ZNDYn

Thanks to @BorisBernhardt, Nicole Eichert, @hong_seok_jun, @acehigh1952, @TheNeuro and many other great discussions! For more insight on hippocampal evolution and organization, check out our recent studies https://t.co/9S3zGuL6CB https://t.co/IZkv3kHUvn

So, latent space predictive coding over multiple timescales may not be all-you-need, but it may be what-you-need

It's possible (IMO probable) for multiple optimizers to act on the same substrate. Reinforcement learning, and perhaps many others, may act on a NN at the same time as latent time-equivariant prediction. Multiple optimizers are actually needed to avoid representational collapse!

The short answer is: animals predict many latent future “inputs” using memory over multiple timescales (time-equivariance). This is not unlike MANNs or self-attention to memory discussed by @jcrwhittington and @behrenstimb ( https://t.co/qU0xtkC5id)

Crossing brain evolution, cognitive psychology, artificial intelligence, and own framework of hippocampal organization, we try to explain why animals are so conspicuously fast at unstructured learning while AIs still take millions of samples + expressive feedback

I’ve been a hippocampus man my entire research career 🍤🧠👨‍🔬. When I first read @ylecun's “JEPA” white paper ( https://t.co/c2XGC0T9CK), I knew I’d found a cornerstone. I’m pleased to finally share my preprinted Perspective! https://t.co/nujQaI0SRu

Why do varied DNN designs yield equally good models of human vision? Our preprint with @michaelfbonner shows that diverse DNNs represent images with a shared set of latent dimensions, and these shared dimensions turn out to also be the most brain-aligned. https://t.co/vtOOYHQb47

Finally out! 🧠🍤 Differential reorganization of episodic and semantic memory systems in epilepsy-related mesiotemporal pathology https://t.co/UvgOrqS8z9 @BorisBernhardt @J_Royer_ @KeXie20 @jordandekraker @rcruces2 @CaciagliLorenzo @the_mindwanders @ale

Very strong opening talk at #CogSci2024 by @morganbarense on taking memory research outside of the lab. So much food for thought. @cogsci_soc

New layer-fMRI preprint investigating the venous vascular bias across subfields of Hippocampus. By @VPf4ffenrot et al., https://t.co/sjfhNKGLM9

Did someone say "Hippocampal laminae" 👀👀👀

🔊Announcement🔊! The @neuroak & @jclauneuro labs are 𝙝𝙞𝙧𝙞𝙣𝙜 𝙩𝙬𝙤✌️𝙧𝙚𝙨𝙚𝙖𝙧𝙘𝙝 𝙨𝙤𝙛𝙩𝙬𝙖𝙧𝙚 𝙙𝙚𝙫𝙚𝙡𝙤𝙥𝙚𝙧𝙨 in basic + translational neuroimaging @WesternU.🧠🔭 Supported by @TRIDENTPCT @Brains_CAN. https://t.co/dPheXyh4RS [Job ID: 35643 + 35752] #ldnont 🇨🇦

So impressed with @jordandekraker's poster at #CNS2024 on "Hippomaps", a superb new toolbox and online repository for the mapping and contextualization of the hippocampal data in the human brain. You can learn more about their pioneering work here: https://t.co/O2f1nG6a8S

Check out HippoMaps at https://t.co/JIkKRcuTHC with @BorisBernhardt @acehigh1952 @donna_gift @J_Royer_ @neuroak @BradKarat @sofievalk @JustineYHansen @misicbata @MarkusAxer @TheNeuro_MNI et al

Thanks to @BorisBernhardt, @acehigh1952, @neuroak, @TheNeuro_MNI, @fz_juelich, @eLife , and many others!

HOWEVER, we still saw some modest improvements in unfolded feature registration in MRI (sharper averages):

So can I get 0.5mm subfield registration in my 1mm T1w MRI data? Not necessarily: 1) Ground-truth is not straightforward to determine in such cases 2) Thickness, gyrification, and curvature will be less detailed in a standard scan, making their registration less impactful

The new unfolded-registration protocol and subfield atlas are now fully implemented in the latest HippUnfold: https://t.co/iEpcaLgVbv