A 𝙝𝙪𝙜𝙚 𝙩𝙝𝙖𝙣𝙠𝙨 to my supervisors & co-authors @cianodonnell.bsky.social , John Wade & @conjh.bsky.social !

We are excited for what these principles imply both for understanding the brain & for designing more robust AI.

Lyapunov dimension of dynamics

𝗧𝗵𝗲 𝘁𝗿𝗮𝗱𝗲-𝗼𝗳𝗳: The efficiency & robustness comes with a price. We found biological wiring leads to lower-dimensional, less expressive dynamics.

We find that neuron self-recurrency is a key feature which stabilises spectral properties & dynamics, inducing robustness to node loss & hyperparameter variation.

Relative spectral radius under random pruning

𝗥𝗼𝗯𝘂𝘀𝘁𝗻𝗲𝘀𝘀: CoNNs also proved to be significantly more resilient to:
- node loss, maintaining performance & spectral radius under pruning longer than conventional random networks
- parameter perturbation, where CoNNs better maintained critical dynamics & smoother dynamical transitions.

Crucially, we found that specific structural features (like self-loops, clustering, and node degree) correlated with task engagement, meaning the network's physical structure directly shapes how neurons share the workload!

Participation ratios of neural engagement distributions

𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆: We found that CoNNs generally achieved comparable task performance to classic Echo State Networks, but with lower “wiring costs” (smaller weights) & lower “energy demands” (reduced neural activity).

CoNNs also showed more specialised engagement (smaller subsets of contributing neurons).

Constructing connectome-based neural networks

Really excited to share a new preprint!

𝗛𝗼𝘄 𝗱𝗼 𝗯𝗿𝗮𝗶𝗻𝘀 𝘀𝘁𝗮𝘆 𝗿𝗼𝗯𝘂𝘀𝘁 & 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝘄𝗵𝗶𝗹𝗲 𝗯𝗲𝗶𝗻𝗴 𝗶𝗻𝗰𝗿𝗲𝗱𝗶𝗯𝗹𝘆 𝘀𝗽𝗮𝗿𝘀𝗲? We explore how!
www.biorxiv.org/content/10.6...

We built Connectome-based Neural Networks (CoNNs) using Drosophila wiring (larva&adult) & compared with random networks with same sparsity.

Excited to read this!

Non-random brain connectome wiring enables robust and efficient neural network function under high sparsity

www.biorxiv.org/content/10.6...

Non-random brain connectome wiring enables robust and efficient neural network function under high sparsity www.biorxiv.org/content/10.64898/2026.03...

New preprint from the lab on synapse development in the nascent neocortical hierarchy!

Using mice that label MAGUK proteins developed by Seth Grant we find key differences in the laminar maturation of association and sensory motor cortices, including delayed, cortex-wide maturation of L1 synapses.

Region- and layer-specific glutamatergic synapse development in the nascent cortical hierarchy www.biorxiv.org/content/10.64898/2026.02...

Will this be available in the UK or EU? :)

Photo of James McAllister from Ulster University giving a presentation on his work on brain-inspired reservoir computing

Lovely time today at the 2nd All-Ireland Computational Neuroscience Symposium at Queens U Belfast. It's a small but growing community on the island. imo the talks were all genuinely great and I'm really excited for the future

Round 3 back in Dublin next August apparently if anyone wants to join! ☘️🧠

This is so fab! What a brilliant resource, thanks John!

Very well done!

@camneuro.bsky.social our amazing PhD student James McAllister is giving a talk at 2pm in CBL on drosophila-larva connectome-based recurrent neural networks
talks.cam.ac.uk/talk/index/2...

This makes me very happy!

Amazing, thank you so much! And thanks for such an awesome network tool.

Thank you!

I was using graph-tool (it's fabulous btw!) in Colab a month or so ago and it worked perfectly, but it now is complaining at the part "from graph_tool.all import *" saying "ModuleNotFoundError: No module named 'graph_tool'"...is there something that hasn't yet been updated in Colab? Thanks!

This is a great write-up of a brilliant paper! :)

Schematic diagram of reservoir computing neural network, with drosophila larva connectome overlaid as reservoir connectivity.

Delighted @jajmca.bsky.social's abstract was accepted for COSYNE 2025, for his work on linking neural circuit structure to function via drosophila-connectome-based reservoir computing. With @conjh.bsky.social, John Wade and myself.

I wouldn't have thought to work out the probability (either as a frequentist or Bayesian) of being Rick rolled by the first footnote!! 😂