@lcparra.bsky.social
NeuroAI, applied neuroscience, machine learning, brain stimulation, neural signals, medical imaging - physicist by training. parralab.org
"The BBC World Service has compiled material on over 160 cases where children have been shot in Gaza, and found that in 95 cases the child was shot in the head or chest.
In most of these cases the victim was under 12 years old."
www.youtube.com/watch?v=TjP6...
Interestingly, they chose to keep the dynamic A constant. Friston et al. assume A to be modulated by the task. The assumption that the brain has (largely) constant dynamic is consistent with our findings in iEEG
elifesciences.org/articles/104...
We are finding the same in EEG. Preprint coming ...
Nice work modeling behavior and EEG as dynamical systems. The State space model is similar to the "dynamical causal model" (without all the fantastical claims). Their code seems to work well in high dimensions, whereas DCM has only been used in small problems, despite both using the EM algorithm.
28.07.2025 04:15 — 👍 6 🔁 1 💬 1 📌 0The word “Research” is doing way too much work. We need separate words for “creating new verifiable knowledge” and “looking shit up on the internet”
25.07.2025 15:58 — 👍 597 🔁 132 💬 21 📌 11
How reliable is OPM-MEG, and how does it compare to other neuroimaging modalities? 🤔
In a new preprint with @s-michelmann.bsky.social, we evaluate the reliability of OPM-MEG within & between individuals, and compare it to fMRI & iEEG during repeated movie viewing. 🧠
📄 doi.org/10.1101/2025...
Here is another idea from biology on how to make ANNs learn more efficiently during continual learning. Adjust weights in first layers early in development [to learn shared representations] and keep higher layers plastic for longer [to learn new tasks].
www.biorxiv.org/content/10.1...
I would go further, don't use EEG for localization. Why? Because the mixing matrix is not invertible. If you try it anyway, all you will get back is the assumptions you put in.
16.07.2025 14:03 — 👍 1 🔁 0 💬 1 📌 0
The problem with "prediction error encoding" is the part about "encoding".
But there is a ton of evidence for prediction error or novelty modulating sensory responses, e.g.
while people watch movies the entire brain seems to be modulated by surprise:
www.nature.com/articles/s41...
🧪 After the workshop on 7/7, the FDA and NIH are asking for public comments on implementing novel methodologies and other strategies to reduce the use of animal testing by 7/14:
www.fda.gov/news-events/...
Together with @repromancer.bsky.social, I have been musing for a while that the exponentiated gradient algorithm we've advocated for comp neuro would work well with low-precision ANNs.
This group got it working!
arxiv.org/abs/2506.17768
May be a great way to reduce AI energy use!!!
#MLSky 🧪
Do you know how crazy the weather has been? It is about to get crazier.
www.intellinews.com/southern-oce...
Clever and simple idea of how to make vision models see the "big picture" rather than relying on coincidental details in an image: Train it first with blurry, colorless, low-contrast images, and only gradually improve them during development. Apparently that is how humans learn to see. So cool!
09.07.2025 10:43 — 👍 2 🔁 0 💬 0 📌 0Congress is trying to kill new investment in solar and wind power.
Meanwhile global warming is accelerating.
This is a death wish, on a national and global scale.
It seems to me the solution is to model a ramp as a ramp. Both you and @spk3lly.bsky.social seem to agree there is a ramp. So why not model that? It's all just linear regression, whether you take averages (as Simon does), fit impulse responses (as you do) or fit ramps (anyone yet?).
29.06.2025 01:31 — 👍 0 🔁 0 💬 1 📌 0
Color indicates if studies make code and trained weights openly available (red: not available, orange: upon request, green: no restriction)
We developed new state-of-the-art AI to detect breast cancer in MRI. This foundation model was trained on 30,000 MRI exams.
Paper: doi.org/10.1148/ryai...
Code: github.com/lkshrsch/Bre...
I still use this as an example of storytelling with visualization when I teach #datavis
24.06.2025 16:41 — 👍 24 🔁 7 💬 1 📌 0
A breakup letter to ChatGPT
www.nytimes.com/2025/06/18/l...
No Kings in Kingston, NY #NoKingsDay
14.06.2025 19:59 — 👍 5 🔁 1 💬 0 📌 1
"Generative A.I. chatbots are going down conspiratorial rabbit holes and endorsing wild, mystical belief systems. For some people, conversations with the technology can deeply distort reality."
www.nytimes.com/2025/06/13/t...
Routine immunization of U.S. children born 1994–2018 will prevent an estimated 419 million illnesses, 936,000 deaths, and $1.9 trillion in societal costs over their lifetimes.
Newer vaccines like HPV are on track to reduce cervical cancers by tens of thousands of cases...
😂😂 I read this many times, and always read it as "altered". My internal dynamic overpowered the stimulus. Sorry about that. You were paying close attention. 🙏
11.06.2025 00:19 — 👍 1 🔁 0 💬 0 📌 0
Evidence in support of the "Brain from the inside out":
Response to movies is dominated by intrinsic brain dynamics. The dynamic itself is barely alternatered by the task of watching movies.
Human intracranial recordings from 5000 electrodes, now out in eLife.
elifesciences.org/reviewed-pre...
A graph showing NIH funding for new and competitive renewal grants for FY2024 and FY2025 showing a much lower rate of investment for FY2025.
I updated my funding curves last night.
Here are the results for new and competitive renewal grants.
The rate of investment has dropped substantially. This is apparently due to a requirement for DOGE review.
I am calling this effect the "Downward DOGE"
1/2
I’m filled with overwhelming dread at Google’s launch of Veo 3 in Gemini. The amount of harm this AI generator will cause can’t be overstated.
If it’s online it’s not to be trusted. Don’t believe your eyes or your ears.
Don’t believe me? I’ll post some of Veo3 AI videos in this thread.
The intrinsic dynamic of the brain dominates its response to external stimulation. ... Think of waves on a quiet lake in response to pebbles you throw in.
elifesciences.org/reviewed-pre...
1/6 Why does the brain maintain such precise excitatory-inhibitory balance?
Our new preprint explores a provocative idea: Small, targeted deviations from this balance may serve a purpose: to encode local error signals for learning.
www.biorxiv.org/content/10.1...
led by @jrbch.bsky.social
My latest: A deep dive into using Claude Code to generate a tool for review of scientific papers. russpoldrack.substack.com/p/reviewing-...
25.05.2025 15:24 — 👍 18 🔁 3 💬 0 📌 0
🚨Hiding under 6 sections of science sounding rhetoric is an edict to appoint officers at each agency that will be responsible for detecting research that isn't "Gold Standard Science" as defined by the White House🚨
Stay tuned for rapid response action!
www.whitehouse.gov/presidential...
All of NIH funding to Northwestern University is frozen. This pause includes noncompeting approved funding, new and competing grants with fundable scores. No reimbursements for money already spent have been received since March. This situation is rarely reported so please Please get the word out!
23.05.2025 17:56 — 👍 2286 🔁 1340 💬 35 📌 32
The "big, beautiful" federal budget bill passed in the House while you were sleeping includes 1 trillion dollars - that's ONE TRILLION - in cuts to health and human service programs, including Medicaid and SNAP, so they can offer $1 trillion in tax cuts for those who make more than $500,000.
1/3
