arnau-m-l

Should neuroscientists ‘vibe code’? Researchers are developing software entirely through natural language conversations with advanced large language models. The trend is transforming how research gets done—but it also presents new…

An in-depth perspective piece discussing the transformative impact of AI coding tools in neuroscience research, featuring my work on SpikeAgent and AI interfaces for neural data analysis. @zuwan-lin.bsky.social
www.thetransmitter.org/craft-and-ca...

Riemannian Geometry for the Classification of Brain States with Intracortical Brain Recordings Geometric machine learning is applied to decode brain states from invasive intracortical neural recordings, extending Riemannian methods to the invasive regime where data is scarcer and less stationa...

Riemannian Geometry for the Classification of Brain States with Intracortical Brain Recordings - Marin‐Llobet - Advanced Intelligent Systems - Wiley Online Library advanced.onlinelibrary.wiley.com/doi/full/10....

GitHub - arnaumarin/LFP-Riemannian: Riemannian for Brain Dynamics Riemannian for Brain Dynamics. Contribute to arnaumarin/LFP-Riemannian development by creating an account on GitHub.

Let us know if you have any thoughts!

Code: github.com/arnaumarin/L...

Riemannian Geometry for the Classification of Brain States with Intracortical Brain Recordings Geometric machine learning is applied to decode brain states from invasive intracortical neural recordings, extending Riemannian methods to the invasive regime where data is scarcer and less stationa...

New paper alert!

In here, we (@ldallap.bsky.social , @mavi_sanchez and co) tap the intrinsic geometry of data to classify brain states on Riemannian manifolds! A lightweight and more INTERPRETABLE alternative to conventional deep learning architectures! Paper: doi.org/10.1002/aisy...

Riemannian Geometry for the Classification of Brain States with Intracortical Brain Recordings Geometric machine learning is applied to decode brain states from invasive intracortical neural recordings, extending Riemannian methods to the invasive regime where data is scarcer and less stationa...

Now out in Advanced Intelligent Systems: Exploiting underlying data geometry, we classify brain states efficiently using Riemannian manifolds. A lightweight and interpretable alternative to DNNs. Spearheaded by
@arnauya.bsky.social!

Check it out: advanced.onlinelibrary.wiley.com/doi/10.1002/...

An AI Agent for cell-type specific brain computer interfaces https://www.biorxiv.org/content/10.1101/2025.09.11.675660v1

Should neuroscientists ‘vibe code’? Researchers are developing software entirely through natural language conversations with advanced large language models. The trend is transforming how research gets done—but it also presents new…

An in-depth perspective piece discussing the transformative impact of AI coding tools in neuroscience research, featuring my work on SpikeAgent and AI interfaces for neural data analysis. @zuwan-lin.bsky.social
www.thetransmitter.org/craft-and-ca...

Great work! I realized there was a Zenodo link but I can't find it. How can I locate it?

An autonomous AI agent for universal behavior analysis Behavior analysis across species represents a fundamental challenge in neuroscience, psychology, and ethology, typically requiring extensive expert knowledge and labor-intensive processes that limit r...

We are very excited to share a preprint for our BehaveAgent, the first fully autonomous AI agent for universal behavior analysis. BehaveAgent is the result of rewarding work with ‪Zuwan Lin and the amazing team in JiaLiu's lab at @harvard.edu . 1/6

www.biorxiv.org/content/10.1...

Realigning representational drift in mouse visual cortex by flexible brain-machine interfaces The ability to stably decode brain activity is crucial for brain-machine interfaces (BMIs), which are often compromised by recording instability due to immune responses and probe drifting. In addition...

www.biorxiv.org/content/10.1...

Neural manifolds: more than the sum of their neurons - Nature Reviews Neuroscience In this Journal Club, Juan Gallego discusses a 2014 article that provided a first causal hint that neural manifolds may not only be a convenient way to interpret neural population activity.

Nature Reviews Neuroscience

Neural manifolds: more than the sum of their neurons

www.nature.com/articles/s41...

I'm not entirely sure I understand what you mean - but essentially, in the paper we found that by making it a dual-classification problem (+ the clustering with high-confidence samples) we could refine the classifications of MA states (which, can be challenging to distinguish with the other states)

GitHub - arnaumarin/LFPDeepStates: neural models for anesthesia stage transition classification neural models for anesthesia stage transition classification - arnaumarin/LFPDeepStates

Check out our fully-open source code below!
github.com/arnaumarin/L...

Neural models for detection and classification of brain states and transitions - Communications Biology A deep learning self-supervised hybrid CNN-autoencoder model is used to detect brain states and transitions, like wakefulness, slow oscillations and microarousals, during the emergence from anesthesia...

A deep learning self-supervised hybrid CNNautoencoder model is used to detect brain states and transitions, like wakefulness, slow oscillations and microarousals, during the emergence from anesthesia in cortical local field potentials
doi.org/10.1038/s420...

Neural models for detection and classification of brain states and transitions Communications Biology - A deep learning self-supervised hybrid CNN-autoencoder model is used to detect brain states and transitions, like wakefulness, slow oscillations and microarousals, during...

rdcu.be/ehyfl
Finally we see our work published! w/ Arnau Manasanch, @ldallap.bsky.social and Mavi Sanchez-Vives!

Llobet, et al.: Riemannian Geometry for the classification of brain states with intracortical brain-computer interfaces https://arxiv.org/abs/2504.05534 https://arxiv.org/pdf/2504.05534 https://arxiv.org/html/2504.05534

Spike sorting AI agent Spike sorting is a fundamental process for decoding neural activity, involving preprocessing, spike detection, feature extraction, clustering, and validation. However, conventional spike sorting metho...

Agents meet neuroscience.

www.biorxiv.org/content/10.1...

GitHub - LiuLab-Bioelectronics-Harvard/SpikeAgent: SpikeAgent is a multimodal LLM-based AI agent that automates and standardizes the spike sorting pipeline SpikeAgent is a multimodal LLM-based AI agent that automates and standardizes the spike sorting pipeline - LiuLab-Bioelectronics-Harvard/SpikeAgent

github.com/LiuLab-Bioel...
!!!!

Check out our last paper on completely automatizing spike sorting using AI agents!
Not only makes the whole process human-free, but also provides an additional layer of information and interpretation that is completely unachievable by any other methods..

Link to the GitHub below!!!

No More Adam: Learning Rate Scaling at Initialization is All You Need In this work, we question the necessity of adaptive gradient methods for training deep neural networks. SGD-SaI is a simple yet effective enhancement to stochastic gradient descent with momentum (SGDM...

This paper looks interesting - it argues that you don’t need adaptive systems like Adam to get good gradient-based training, instead you can just set a learning rate for different groups of units based on initialization:

arxiv.org/abs/2412.11768

#MLSky #NeuroAI

also here!!!
Beniaguev, D., Segev, I., & London, M. (2021). Single cortical neurons as deep artificial neural networks. Neuron, 109(17), 2727-2739.

Best quote from the trailer " I admit that I literally really understand nothing of this"

not necessarily the 'universal computation' but at least its a common language across subjects in neuro. I mean, we can't be sure we record the single-cell resolution same neurons from our very limited (and very downsampled) number of neurons.

𝗪𝗵𝗮𝘁 𝗶𝗻 𝘁𝗵𝗲 𝗯𝗿𝗮𝗶𝗻 𝗶𝗺𝗽𝗹𝗲𝗺𝗲𝗻𝘁𝘀 𝘂𝗻𝗶𝘃𝗲𝗿𝘀𝗮𝗹 𝗰𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻?
Maybe not neurons, perhaps RNA??
Great episode of Brain Inspired with Hessam Akhlaghpour.
Do dynamical systems and/or RNNs implement universal computation? This discussion alone would be worth having.
#neuroscience
braininspired.co/podcast/199/

Adaptive Intelligence: leveraging insights from adaptive behavior in animals to build flexible AI systems Biological intelligence is inherently adaptive -- animals continually adjust their actions based on environmental feedback. However, creating adaptive artificial intelligence (AI) remains a major chal...

Alright #neuroAI 😉 crew and those in #AI #ML perhaps interested in brain-inspired models and what’s next (not just foundation models please!). Here is my vision/roadmap of the next steps. A reflection & projection wrapped into a Perspective. Feedback very much welcomed. 🧠🧪👩‍🔬

arxiv.org/abs/2411.15234

Juan! Love the list so far. I’d love to join if there’s room still :)

I *just* made a neurotech industry starter pack and added you to it.

Looks like it’s all organic so far due to the cool biohybrid work you just dropped!

go.bsky.app/N2LreWi

i was wondering the same