Preview of the talk I'm giving on Friday. #neuroscience #CompNeuro #ComputationalNeuroscience
Recent searches
Search options
#compneuro
0 posts0 participants0 posts today
Low stakes pet peeve of the day: spiking neural network people stop saying SNNs are the third generation of ANNs. They predate them! #compneuro
I'm giving an online talk starting in 15m (as part of UCL's NeuroAI series).
It's on neural architectures and our current line of research trying to figure out what they might be good for (including some philosophy: what might an answer to this question even look like?).
Sign up (free) at this link to get the zoom link:
EventbriteUCL NeuroAI Talk SeriesA series of NeuroAI themed talks organised by the UCL NeuroAI community. Talks will continue on a monthly basis.
Come along to my (free, online) UCL NeuroAI talk next week on neural architectures. What are they good for? All will finally be revealed and you'll never have to think about that question again afterwards. Yep. Definitely that.
Wed 12 Feb 2025 
2-3pm GMT
Details and registration: https://www.eventbrite.co.uk/e/ucl-neuroai-talk-series-tickets-1216638381149
EventbriteUCL NeuroAI Talk SeriesA series of NeuroAI themed talks organised by the UCL NeuroAI community. Talks will continue on a monthly basis.
Visual cognition in multimodal large language models
#CompNeuro #NeuroAI
www.nature.com/articles/s42...
Visual cognition in multimodal...
Bluesky SocialBluesky
What's the right way to think about modularity in the brain? This devilish 
question is a big part of my research now, and it started with this paper with @GabrielBena finally published after the first preprint in 2021!
https://www.nature.com/articles/s41467-024-55188-9
We know the brain is physically structured into distinct areas ("modules"?). We also know that some of these have specialised function. But is there a necessary connection between these two statements? What is the relationship - if any - between 'structural' and 'functional' modularity?
TLDR if you don't want to read the rest: there is no necessary relationship between the two, although when resources are tight, functional modularity is more likely to arise when there's structural modularity. We also found that functional modularity can change over time! Longer version follows.
NatureDynamics of specialization in neural modules under resource constraints - Nature CommunicationsThe extent to which structural modularity in neural networks ensures functional specialization remains unclear. Here the authors show that specialization can emerge in neural modules placed under resource constraints but varies dynamically and is influenced by network architecture and information flow.
Discovered the COSYNE 2025 workshop program is out! 
Looking forward to fascinating computational neuroscience talks & discussions (March 31-April 1, 2025).
Check out the program: https://www.cosyne.org/workshops-program-2025
See you there! #COSYNE2025 #CompNeuro @CosyneMeeting
@CosyneMeeting #Cosyne
COSYNEProgram — COSYNE
New preprint! With Swathi Anil and @marcusghosh.
If you want to get the most out of a multisensory signal, you should take it's temporal structure into account. But which neural architectures do this best? 
bioRxiv · Fusing multisensory signals across channels and timeAnimals continuously combine information across sensory modalities and time, and use these combined signals to guide their behaviour. Picture a predator watching their prey sprint and screech through a field. To date, a range of multisensory algorithms have been proposed to model this process including linear and nonlinear fusion, which combine the inputs from multiple sensory channels via either a sum or nonlinear function. However, many multisensory algorithms treat successive observations independently, and so cannot leverage the temporal structure inherent to naturalistic stimuli. To investigate this, we introduce a novel multisensory task in which we provide the same number of task-relevant signals per trial but vary how this information is presented: from many short bursts to a few long sequences. We demonstrate that multisensory algorithms that treat different time steps as independent, perform sub-optimally on this task. However, simply augmenting these algorithms to integrate across sensory channels and short temporal windows allows them to perform surprisingly well, and comparably to fully recurrent neural networks. Overall, our work: highlights the benefits of fusing multisensory information across channels and time, shows that small increases in circuit/model complexity can lead to significant gains in performance, and provides a novel multisensory task for testing the relevance of this in biological systems.
Key Points
### Competing Interest Statement
The authors have declared no competing interest.
Emergence and maintenance of modularity in neural networks with Hebbian and anti-Hebbian inhibitory STDP
with R. Bergoin. @gzamora-lopez.bsky.social , G. Deco, M. Quoi
#neuroscience #compneuro #brain #neuralnetwork
#neurodon @virati @hnp_geneva
An open access paper :
Gamma oscillatory complexity conveys behavioral information in hippocampal networks with V Douchamps, M di Volo, D Battaglia and R Goutagny
Nat Comm 15, 1849 (2024)
Our findings challenge the idea of rigid gamma sub-bands, showing that behavior shapes ensembles of irregular gamma elements that evolve with learning and depend on hippocampal layers
https://www.nature.com/articles/s41467-024-46012-5
#neuroscience #compneuro #brain #neuralnetwork
#neurodon @virati @hnp_geneva
Continued thread
(10/n) If you’ve made it this far, you’ll definitely want to check out the full paper. Grab your copy here:
https://www.biorxiv.org/content/10.1101/2024.12.17.628339v1
Sharing is highly appreciated!
#compneuro #neuroscience #NeuroAI #dynamicalsystems
bioRxiv · From spiking neuronal networks to interpretable dynamics: a diffusion-approximation frameworkModeling and interpreting the complex recurrent dynamics of neuronal spiking activity is essential to understanding how networks implement behavior and cognition. Nonlinear Hawkes process models can capture a large range of spiking dynamics, but remain difficult to interpret, due to their discontinuous and stochastic nature. To address this challenge, we introduce a novel framework based on a piecewise deterministic Markov process representation of the nonlinear Hawkes process (NH-PDMP) followed by a diffusion approximation. We analytically derive stability conditions and dynamical properties of the obtained diffusion processes for single-neuron and network models. We established the accuracy of the diffusion approximation framework by comparing it with exact continuous-time simulations of the original neuronal NH-PDMP models. Our framework offers an analytical and geometric account of the neuronal dynamics repertoire captured by nonlinear Hawkes process models, both for the canonical responses of single-neurons and neuronal-network dynamics, such as winner-take-all and traveling wave phenomena. Applied to human and nonhuman primate recordings of neuronal spiking activity during speech processing and motor tasks, respectively, our approach revealed that task features can be retrieved from the dynamical landscape of the fitted models. The combination of NH-PDMP representations and diffusion approximations thus provides a novel dynamical analysis framework to reveal single-neuron and neuronal-population dynamics directly from models fitted to spiking data.
### Competing Interest Statement
The authors have declared no competing interest.
We have a new preprint on the emergence of orientation selectivity in layers 2/3 and 4 of the mouse. We use data from the Allen Institute's Microns project, which includes structure plus function of thousands of neurons, to constrain network models that account for the observations and hint some key features on the origin of tuning in L2/3. For any feedback, do not hesitate to contact us!
bioRxiv · Connectome-based models of feature selectivity in a cortical circuitFeature selectivity, the ability of neurons to respond preferentially to specific stimulus configurations, is a fundamental building block of cortical functions. Various mechanisms have been proposed to explain its origins, differing primarily in their assumptions about the connectivity between neurons. Some models attribute selectivity to structured, tuning-dependent feedforward or recurrent connections, whereas others suggest it can emerge within randomly connected networks when interactions are sufficiently strong. This range of plausible explanations makes it challenging to identify the core mechanisms of feature selectivity in the cortex. We developed a novel, data-driven approach to construct mechanistic models by utilizing connectomic data-synaptic wiring diagrams obtained through electron microscopy-to minimize preconceived assumptions about the underlying connectivity. With this approach, leveraging the MICrONS dataset, we investigate the mechanisms governing selectivity to oriented visual stimuli in layer 2/3 of mouse primary visual cortex. We show that connectome-constrained network models replicate experimental neural responses and point to connectivity heterogeneity as the dominant factor shaping selectivity, with structured recurrent and feedforward connections having a noticeable but secondary effect in its amplification. These findings provide novel insights on the mechanisms underlying feature selectivity in cortex and highlight the potential of connectome-based models for exploring the mechanistic basis of cortical functions.
### Competing Interest Statement
The authors have declared no competing interest.
My latest work “Zero-shot counting with a dual-stream neural network model” is now published in Neuron. We present evidence for an enactive theory of the role of posterior parietal cortex in visual scene understanding and structure learning. Like the primate brain, our model apprehends a visual scene via a sequence of foveated glimpses. Both glimpse contents and glimpse locations are fed into our model, enabling the model to learn abstractions that are grounded in action (here, eye movements) rather than merely in the sensory domain. We show that this architecture enables zero-shot generalization of a previously learned structure (numerosity) to new objects in new contexts in a setting where a vanilla CNN fails to generalize. Our model also replicates several signatures of (the development of) human counting behaviour and learns representations that mimic neural codes for space and number in the primate brain.
#neuralnetworks #compneuro #neuroscience #4ECognition #attention
https://www.sciencedirect.com/science/article/pii/S0896627324007293
Continued thread
I've added a new chapter on #VariationalAutoencoders (VAE), including an exercise that shows how to train a #VAE to predict behavior from neural data inputs.
We just completed a new course on #DimensionalityReduction in #Neuroscience, and the full teaching material 
is now freely available (CC BY 4.0 license):
https://www.fabriziomusacchio.com/blog/2024-10-24-dimensionality_reduction_in_neuroscience/
The course is designed to provide an introductory overview of the application of dimensionality reduction techniques for neuroscientists and data scientists alike, focusing on how to handle the increasingly high-dimensional datasets generated by modern neuroscience research.
Functional responses in the brain to linguistic inputs are spatially organized -- but why? We show that a simple smoothness loss added to language model training explains a range of topographic phenomena in neuroscience: https://arxiv.org/abs/2410.11516
#NeuroAI #neuroscience #compneuro #language
PhD position openings in my group! Check out topics and how to apply below. At the moment, I'm particularly interested in the topic of modularity in both biological and artificial networks, and how it can be used to scale up intelligent processes.
https://neural-reckoning.org/openings.html
Note that competition for PhD funding at Imperial is pretty strong these days, so worth checking if your undergrad/masters grades are equivalent to UK 1st/Distinction before applying. I'm afraid I don't take self-funded PhD students.
neural-reckoning.orgJoin us
Replied in thread
Oh Tx A!
What a pleasure reading classic pioneers humble candid remarks!
Contrast to nowadays endless new-papers new-models proposals without any scientific model self awareness on bias towards invalidation, and lacking such kind of remarks of owns models testable limitations, with such candid remarks alerts to fellow readers!
Also from B. Katz, the other coauthor in one of the #HodgkinHuxley original quintet of seminal #compneuro papers!
Reminds of their 1952b that remarks:
I'm not the physicist here, but honestly, framed as "showing how statistical physics can explain and implement forms of learning" sounds great as a Nobel Prize motive.
Now get with me through the rabbit hole...
(Maybe I should flip the edges and put something else in the center of the meme)
Belief stickiness and poor state inference aka obsessions are anticorrelated with plasma levels of SSRI esticalopram in a randomized double-blind placebo controlled study #neuroscience #compneuro #ComputationalPsychiatry https://www.biorxiv.org/content/10.1101/2023.12.08.570769v1
bioRxiv · Serotonin Reduces Belief StickinessSerotonin fosters cognitive flexibility, but how, exactly, remains unclear. We show that serotonin reduces belief stickiness: the tendency to get “stuck” in a belief about the state of the world despite incoming contradicting evidence. Participants performed a task assessing belief stickiness in a randomized, double-blind, placebo-controlled study using a single dose of the selective serotonin reuptake inhibitor (SSRI) escitalopram. In the escitalopram group, higher escitalopram plasma levels reduced belief stickiness more, resulting in better inference about the state of the world. Moreover, participants with sufficiently high escitalopram plasma levels had less belief stickiness, and therefore better state inference, than participants on placebo. Exaggerated belief stickiness is exemplified by obsessions: “sticky” thoughts that persist despite contradicting evidence. Indeed, participants with more obsessions had greater belief stickiness, and therefore worse state inference. The opposite relations of escitalopram and obsessions with belief stickiness may explain the therapeutic effect of SSRIs in obsessive-compulsive disorder.
### Competing Interest Statement
The authors have declared no competing interest.