Andrea Costantino
@costantinoai.bsky.social
👀🧠🤖 cognitive neuroscientist @ Hoplab, KU Leuven | interested in vision and learning
Reposted by Andrea Costantino
Y’all are reading this paper in the wrong way.
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
“Our findings challenge the conventional focus on low-dimensional coding subspaces as a sufficient framework for understanding neural computations, demonstrating that dimensions previously considered task-irrelevant and accounting for little variance can have a critical role in driving behavior.”
Neural dynamics outside task-coding dimensions drive decision trajectories through transient amplification
Most behaviors involve neural dynamics in high-dimensional activity spaces. A common approach is to extract dimensions that capture task-related variability, such as those separating stimuli or choice...
www.biorxiv.org
November 25, 2025 at 4:16 PM
Y’all are reading this paper in the wrong way.
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
We love to trash dominant hypothesis, but we need to look for evidence against the manifold hypothesis elsewhere:
This elegant work doesn't show neural dynamics are high D, nor that we should stop using PCA
It’s quite the opposite!
(thread)
Huh, funny. 40 years passed, but this is still so relatable
November 25, 2025 at 8:28 AM
Huh, funny. 40 years passed, but this is still so relatable
Reposted by Andrea Costantino
Not directly related to @neurograce.bsky.social’s comment here but to the thread: I am baffled by the number of people who seem to think the claim is that all brain computation is low-d vs. the *emprical* finding that task-related neural activity is (linear) low-d in many (not all) cases.
This kind of stuff is why I say that I worry that the tools of neuroscience are not properly vetted
“Our findings challenge the conventional focus on low-dimensional coding subspaces as a sufficient framework for understanding neural computations, demonstrating that dimensions previously considered task-irrelevant and accounting for little variance can have a critical role in driving behavior.”
November 24, 2025 at 1:16 AM
Not directly related to @neurograce.bsky.social’s comment here but to the thread: I am baffled by the number of people who seem to think the claim is that all brain computation is low-d vs. the *emprical* finding that task-related neural activity is (linear) low-d in many (not all) cases.
Reposted by Andrea Costantino
1. 🧵 Thread: What happens to the visual brain after early transient blindness?
Our new Nature Communications paper examines a rare population: people born with dense bilateral cataracts—a short blindness occurring during a critical window of visual development.
🔗 rdcu.be/eQjMH
Our new Nature Communications paper examines a rare population: people born with dense bilateral cataracts—a short blindness occurring during a critical window of visual development.
🔗 rdcu.be/eQjMH
November 19, 2025 at 9:07 AM
1. 🧵 Thread: What happens to the visual brain after early transient blindness?
Our new Nature Communications paper examines a rare population: people born with dense bilateral cataracts—a short blindness occurring during a critical window of visual development.
🔗 rdcu.be/eQjMH
Our new Nature Communications paper examines a rare population: people born with dense bilateral cataracts—a short blindness occurring during a critical window of visual development.
🔗 rdcu.be/eQjMH
Reposted by Andrea Costantino
Low-Dimensional and Optimised Representations of High-Level Information in the Expert Brain https://www.biorxiv.org/content/10.1101/2025.11.12.688012v1
November 12, 2025 at 5:16 PM
Low-Dimensional and Optimised Representations of High-Level Information in the Expert Brain https://www.biorxiv.org/content/10.1101/2025.11.12.688012v1
Super excited to share a new preprint!
We asked a simple-but-big question:
What changes in the brain when someone becomes an expert?
Using chess ♟️ + fMRI 🧠 + representational geometry & dimensionality 📈, we ask:
1️⃣ WHAT information is encoded?
2️⃣ HOW is it structured?
3️⃣ WHERE is it expressed?
1/n
We asked a simple-but-big question:
What changes in the brain when someone becomes an expert?
Using chess ♟️ + fMRI 🧠 + representational geometry & dimensionality 📈, we ask:
1️⃣ WHAT information is encoded?
2️⃣ HOW is it structured?
3️⃣ WHERE is it expressed?
1/n
November 12, 2025 at 10:56 PM
Super excited to share a new preprint!
We asked a simple-but-big question:
What changes in the brain when someone becomes an expert?
Using chess ♟️ + fMRI 🧠 + representational geometry & dimensionality 📈, we ask:
1️⃣ WHAT information is encoded?
2️⃣ HOW is it structured?
3️⃣ WHERE is it expressed?
1/n
We asked a simple-but-big question:
What changes in the brain when someone becomes an expert?
Using chess ♟️ + fMRI 🧠 + representational geometry & dimensionality 📈, we ask:
1️⃣ WHAT information is encoded?
2️⃣ HOW is it structured?
3️⃣ WHERE is it expressed?
1/n
Reposted by Andrea Costantino
New paper (and thread) on the representational dynamics of the main dimensions of object space: jov.arvojournals.org/article.aspx... 1/n
Representational dynamics of the main dimensions of object space: Face/body selectivity aligns temporally with animal taxonomy but not with animacy | JOV | ARVO Journals
jov.arvojournals.org
November 9, 2025 at 1:48 PM
New paper (and thread) on the representational dynamics of the main dimensions of object space: jov.arvojournals.org/article.aspx... 1/n
Reposted by Andrea Costantino
It's finally out! Our work with @hansopdebeeck.bsky.social & @costantinoai.bsky.social is published.
We went looking for dissociations between types of recurrence in DNNs, but we found something quite different.. hopefully that can tell us somehting about our models!
rdcu.be/eLwBA
We went looking for dissociations between types of recurrence in DNNs, but we found something quite different.. hopefully that can tell us somehting about our models!
rdcu.be/eLwBA
October 27, 2025 at 4:18 PM
It's finally out! Our work with @hansopdebeeck.bsky.social & @costantinoai.bsky.social is published.
We went looking for dissociations between types of recurrence in DNNs, but we found something quite different.. hopefully that can tell us somehting about our models!
rdcu.be/eLwBA
We went looking for dissociations between types of recurrence in DNNs, but we found something quite different.. hopefully that can tell us somehting about our models!
rdcu.be/eLwBA
Reposted by Andrea Costantino
We’re hiring! We’re looking for two RAs to study neuroplasticity in sight loss, sight rescue and development in children and adults @ucl.ac.uk using a wide range of neuroimaging and behavioral methods. Please help spread the word! Apply by 16 Oct! t.ly/q3aYe #neurojobs #NeuroSkyence
Research Assistant at UCL
Recruiting now: Research Assistant on jobs.ac.uk. Click for details and explore more academic job opportunities on the top job board
t.ly
September 27, 2025 at 10:57 AM
We’re hiring! We’re looking for two RAs to study neuroplasticity in sight loss, sight rescue and development in children and adults @ucl.ac.uk using a wide range of neuroimaging and behavioral methods. Please help spread the word! Apply by 16 Oct! t.ly/q3aYe #neurojobs #NeuroSkyence
Reposted by Andrea Costantino
🧠 New preprint: Why do deep neural networks predict brain responses so well?
We find a striking dissociation: it’s not shared object recognition. Alignment is driven by sensitivity to texture-like local statistics.
📊 Study: n=57, 624k trials, 5 models doi.org/10.1101/2025...
We find a striking dissociation: it’s not shared object recognition. Alignment is driven by sensitivity to texture-like local statistics.
📊 Study: n=57, 624k trials, 5 models doi.org/10.1101/2025...
September 8, 2025 at 6:32 PM
🧠 New preprint: Why do deep neural networks predict brain responses so well?
We find a striking dissociation: it’s not shared object recognition. Alignment is driven by sensitivity to texture-like local statistics.
📊 Study: n=57, 624k trials, 5 models doi.org/10.1101/2025...
We find a striking dissociation: it’s not shared object recognition. Alignment is driven by sensitivity to texture-like local statistics.
📊 Study: n=57, 624k trials, 5 models doi.org/10.1101/2025...
It was great reconnecting with friends and colleagues at #CCN2025 in Amsterdam and presenting our latest #expertise work.
We 👀 into how #chess experts represent the board, and how the content, structure, and location of these repr shift w/ expertise.⬇️
We 👀 into how #chess experts represent the board, and how the content, structure, and location of these repr shift w/ expertise.⬇️
August 19, 2025 at 9:40 PM
It was great reconnecting with friends and colleagues at #CCN2025 in Amsterdam and presenting our latest #expertise work.
We 👀 into how #chess experts represent the board, and how the content, structure, and location of these repr shift w/ expertise.⬇️
We 👀 into how #chess experts represent the board, and how the content, structure, and location of these repr shift w/ expertise.⬇️
Reposted by Andrea Costantino
Neural manifold properties can help us understand how animal brains deal with competing and multifaceted information, execute flexible behaviors and reuse common computations, writes @mattperich.bsky.social.
#neuroskyence
www.thetransmitter.org/neural-dynam...
#neuroskyence
www.thetransmitter.org/neural-dynam...
Neural population-based approaches have opened new windows into neural computations and behavior
Neural manifold properties can help us understand how animal brains deal with complex information, execute flexible behaviors and reuse common computations.
www.thetransmitter.org
August 4, 2025 at 2:56 PM
Neural manifold properties can help us understand how animal brains deal with competing and multifaceted information, execute flexible behaviors and reuse common computations, writes @mattperich.bsky.social.
#neuroskyence
www.thetransmitter.org/neural-dynam...
#neuroskyence
www.thetransmitter.org/neural-dynam...
Reposted by Andrea Costantino
🚨 New preprint alert!
Our latest study, led by @DrewLinsley, examines how deep neural networks (DNNs) optimized for image categorization align with primate vision, using neural and behavioral benchmarks.
Our latest study, led by @DrewLinsley, examines how deep neural networks (DNNs) optimized for image categorization align with primate vision, using neural and behavioral benchmarks.
April 28, 2025 at 1:22 PM
🚨 New preprint alert!
Our latest study, led by @DrewLinsley, examines how deep neural networks (DNNs) optimized for image categorization align with primate vision, using neural and behavioral benchmarks.
Our latest study, led by @DrewLinsley, examines how deep neural networks (DNNs) optimized for image categorization align with primate vision, using neural and behavioral benchmarks.
Reposted by Andrea Costantino
People talk a lot about objects, but what about the softness of a cushion, the greenness of an emerald, or the viscosity of oil? In our work just published @pnas.org, we shed light on how we make sense of the hundreds of materials around us.
www.pnas.org/doi/10.1073/...
www.pnas.org/doi/10.1073/...
March 6, 2025 at 9:36 PM
People talk a lot about objects, but what about the softness of a cushion, the greenness of an emerald, or the viscosity of oil? In our work just published @pnas.org, we shed light on how we make sense of the hundreds of materials around us.
www.pnas.org/doi/10.1073/...
www.pnas.org/doi/10.1073/...
Reposted by Andrea Costantino
You know what I'd love to be able to do?
Research.
Research.
February 13, 2025 at 6:26 PM
You know what I'd love to be able to do?
Research.
Research.
Reposted by Andrea Costantino
Paper🚨 "Objects, Faces, and Spaces: Organizational Principles of Visual Object Perception as Evidenced by Individual Differences in Behavior" by @heidasigurdar.bsky.social & @ingamariao.bsky.social JEP:G editor's choice ->free to read psycnet.apa.org/fulltext/202... #visionscience #psychscisky 🧵1/13
February 13, 2025 at 11:27 AM
Paper🚨 "Objects, Faces, and Spaces: Organizational Principles of Visual Object Perception as Evidenced by Individual Differences in Behavior" by @heidasigurdar.bsky.social & @ingamariao.bsky.social JEP:G editor's choice ->free to read psycnet.apa.org/fulltext/202... #visionscience #psychscisky 🧵1/13
Reposted by Andrea Costantino
🚨PhD opportunity Fall/Winter
2025🚨
Join me in Geneva Switzerland #unige to learn more about colour perception. Using neuroimaging & computational modelling, you'll be working with an international & interdisciplinary team to understand how we transform light into a colourful world!🧠👁️🌈 #neurojobs
2025🚨
Join me in Geneva Switzerland #unige to learn more about colour perception. Using neuroimaging & computational modelling, you'll be working with an international & interdisciplinary team to understand how we transform light into a colourful world!🧠👁️🌈 #neurojobs
January 21, 2025 at 2:08 PM
🚨PhD opportunity Fall/Winter
2025🚨
Join me in Geneva Switzerland #unige to learn more about colour perception. Using neuroimaging & computational modelling, you'll be working with an international & interdisciplinary team to understand how we transform light into a colourful world!🧠👁️🌈 #neurojobs
2025🚨
Join me in Geneva Switzerland #unige to learn more about colour perception. Using neuroimaging & computational modelling, you'll be working with an international & interdisciplinary team to understand how we transform light into a colourful world!🧠👁️🌈 #neurojobs