banner
adeelrazi.bsky.social
Adeel Razi
@adeelrazi.bsky.social
2.7K followers 1.4K following 86 posts
Computational Neuroscientist, NeuroAI, Causality. Monash, UCL, CIFAR. Lab: https://comp-neuro.github.io/
comp-neuro.github.io
Posts Media Videos Starter Packs
Pinned
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · Mar 17
🧵 How do psychedelics shape brain activity?

Our new paper presents the largest neuroimaging study of psilocybin to date—revealing how context structures psychedelic brain states.

Title: Psychedelics Align Brain Activity with Context
Paper: lnkd.in/gt-kMR6d

A thread 🧵👇 1/n
4 40 84
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · 14d
This is BIG. Nobody was expecting this cure to work or even if it was possible ever to treat HD!
1 2 22
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · 14d
Congratulations and looking forward to seeing what you do there!
1
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · 21d
We are hiring for our psilocybin clinical trial.

1. Research Officer (testing and screening), 2 positions
careers.pageuppeople.com/513/cw/en/jo...

2. Research Officer (Peer Navigator)
careers.pageuppeople.com/513/cw/en/jo...

DM if you have questions.

@wellcomeleap.bsky.social
LinkedIn
This link will take you to a page that’s not on LinkedIn
lnkd.in
2 4
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · Sep 2
Thank you very much @prelights.bsky.social and uMontreal Neuro team for coosing and covering our preprint.
prelights.bsky.social
preLights @prelights.bsky.social · Sep 2
Psilocybin restructures brain activity depending on the sensory context in which it is taken

Read this new #preLight by uMontreal Neuro preLighters Loïk Holdrinet, Nour Eltaani and Emma Clini talking about the #preprint of Devon Stoliker, @adeelrazi.bsky.social and the team.
Psychedelics Align Brain Activity with Context - preLights
Psilocybin restructures brain activity depending on the sensory context in which it is taken.
prelights.biologists.com
1 5
Reposted by Adeel Razi
smellosopher.bsky.social
Ann-Sophie Barwich @smellosopher.bsky.social · Aug 27
And we are live!

"Wider Than the Sky: An Alternative to “Mapping” the World Onto the Brain"
A.S. Barwich, S.J. Firestein, M. R. Dietrich
European Journal of Neuroscience

Open Access onlinelibrary.wiley.com/doi/10.1111/...
#NeuroSky #neuro #neuroscience #philsci #hps #sts
Wider Than the Sky: An Alternative to “Mapping” the World Onto the Brain
Challenging the topographic paradigm: The olfactory system's dynamic organization and phenomena like representational drift challenge traditional spatial mapping models, advocating for flexible, r...
onlinelibrary.wiley.com
3 10 40
Reposted by Adeel Razi
misicbata.bsky.social
Bratislav Misic @misicbata.bsky.social · Aug 4
Integrating and interpreting brain maps | doi.org/10.1016/j.ti...

Imaging and recording technologies make it possible to map multiple biological features of the brain. How can these features be conceptually integrated into a coherent understanding of brain structure and function? ⤵️
1 42 72
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · Jul 30
This is for our collaboration with @wellcomeleap.bsky.social on Untangling Addiction.

wellcomeleap.org/ua/program/
Untangling Addiction Program Details | Wellcome Leap: Unconventional Projects. Funded at Scale.
NEW $50M PROGRAMUntangling AddictionWe are pleased to announce the selected performers.Yasmin Hurd, Icahn School of Medicine at Mount SinaiBrett Ginsburg, University of Texas Health Science Center at ...
wellcomeleap.org
2
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · Jul 30
We are hiring for our brand new psilocybin clinical trial. Please DM if you have questions!

1. Senior Research Fellow (Clinical Trial Lead)

careers.pageuppeople.com/513/cw/en/jo...

2. Senior Clinical Psychotherapist

careers.pageuppeople.com/513/cw/en/jo...
1 1 8
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · Jun 1
Very much forward to this @ohbmofficial.bsky.social
ohbmofficial.bsky.social
The Organization for Human Brain Mapping @ohbmofficial.bsky.social · May 28
⭐ Speaker Spotlight: Adeel Razi
🗓️ June 25 | 🕥 10:30 – 11:15
We’re excited to feature Adeel Razi at OHBM 2025.
More information about Adeel in comments!
#OHBM2025
2 7
Reposted by Adeel Razi
stuartoldham.bsky.social
Stuart Oldham @stuartoldham.bsky.social · May 28
Do you like brain network hubs?🧠🌐✳️Do you like genes?🧬What about neurodevelopment?👶What if I told you the latest work by @garedaba.bsky.social and myself combined all of these?🤯🤯🤯

See Gareth's thread for a primer of our findings, then read the paper for the details!
www.biorxiv.org/content/10.1...
13 30
Reposted by Adeel Razi
anayebi.bsky.social
Aran Nayebi @anayebi.bsky.social · May 27
Can a Universal Basic Income (UBI) become feasible—even if AI fully automates existing jobs and creates no new ones?

We derive a closed-form UBI threshold tied to AI capabilities that suggests it's potentially achievable by mid-century even under moderate AI growth assumptions:
2 4 12
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 27
That's really intersting and relevant, will read closely and cite in the related work. We currently cited this one fo binary NNs: arxiv.org/abs/2002.10778
Training Binary Neural Networks using the Bayesian Learning Rule
Neural networks with binary weights are computation-efficient and hardware-friendly, but their training is challenging because it involves a discrete optimization problem. Surprisingly, ignoring the d...
arxiv.org
1
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 27
Reg batchnorm: it's effective in many settings, but can be brittle in others, like when used with small batch sizes, non-i.i.d. data or models with stochasticity in the forward pass. In these cases, the running estimates of mean/variance can drift or misalign with test-time behaviour.

2/2
1
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 27
Yes, absolutely, "noisy" was shorthand & it does depend on the surrogate. What I meant is that common surrogates can have high gradient variance, especially when their outputs saturate. That variance can hurt learning, particularly in deeper networks or those with binary/stochastic activations.
1/2
1 1
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 26
of course, whenever you could!
1
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 26
Paper: arxiv.org/abs/2505.17962

We’d love feedback, extensions, or critiques.

@neuralreckoning.bsky.social @fzenke.bsky.social @wellingmax.bsky.social

#NeuroAI

6/6
A Principled Bayesian Framework for Training Binary and Spiking Neural Networks
We propose a Bayesian framework for training binary and spiking neural networks that achieves state-of-the-art performance without normalisation layers. Unlike commonly used surrogate gradient methods...
arxiv.org
1 2
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 26
Why does KL divergence show up everywhere in machine learning?

Because it's not just a distance, it's the cost of believing your own model too much.

Minimizing KL = reducing surprise = optimizing variational free energy.

A silent principle behind robust inference.

5/6
1 5
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 26
Our key innovation:

- A family of importance-weighted straight-through estimators (IW-ST), which unify and generalize previous methods.
- No need for backprop-through-noise tricks.
- No batch norm.

Just clean, effective training.

4/6
1 1
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 26
We view training as Bayesian inference, minimizing KL divergence between a posterior and an amortized prior.

This lets us derive a principled loss from first principles—grounded in variational free energy, not heuristics.

3/6
1 1
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 26
Binary/spiking neural networks are efficient and brain-inspired—but notoriously difficult to train.

Why? Discrete activations → non-differentiable.

Most current methods either approximate gradients or add noisy surrogates.

We do something different.

2/6
1 3
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 26
🧵Training binary or spiking neural networks is hard.

Gradients vanish, surrogates are noisy, batchnorm is brittle.

We propose a Bayesian approach based on KL divergence minimization—and it works.

Paper: arxiv.org/abs/2505.17962

Work by James Walker & Moein Khajehnejad

@neuro-ai.bsky.social

1/6
A Principled Bayesian Framework for Training Binary and Spiking Neural Networks
We propose a Bayesian framework for training binary and spiking neural networks that achieves state-of-the-art performance without normalisation layers. Unlike commonly used surrogate gradient methods...
arxiv.org
3 12 30
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 25
If brains infer control by predicting their own actions,
should future AI do the same?

Instead of optimizing over actions,
let’s build agents that explain their sensations.

Intelligence may not be about control—but coherence.

#AgencyByInference
1 4
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 25
Maybe intelligence isn’t about maximizing reward…
but minimizing surprise in a world we predictively model.

What if agency is not learned—but inferred?
1 4
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 25
Why do brains rely on inference, uncertainty, and structure…

while AI systems chase rewards in unstructured worlds?

Are we missing something fundamental about how intelligence emerges?

#NeuroAI #InferenceOverOptimization
2 3
adeelrazi.bsky.social
Adeel Razi @adeelrazi.bsky.social · May 15
Yes, but he told us how to quantify it. This was the game changer 😀
1 1