Ole Christian Sylte
@olechsylte.bsky.social
Neuroscientist, postdoc in Stephenson-Jones lab at Sainsbury Wellcome Centre, UCL
Reposted by Ole Christian Sylte
And for me the big question is: What about representational drift? 😄
Manifolds can be mapped, but individual neuronal responses change. That’s neat, but how does it make sense?
How can perception stay stable if the neurons keep changing?
I’m sure manifolds will help us understand that too…
🧠🧪
3/3
Manifolds can be mapped, but individual neuronal responses change. That’s neat, but how does it make sense?
How can perception stay stable if the neurons keep changing?
I’m sure manifolds will help us understand that too…
🧠🧪
3/3
August 4, 2025 at 7:27 PM
And for me the big question is: What about representational drift? 😄
Manifolds can be mapped, but individual neuronal responses change. That’s neat, but how does it make sense?
How can perception stay stable if the neurons keep changing?
I’m sure manifolds will help us understand that too…
🧠🧪
3/3
Manifolds can be mapped, but individual neuronal responses change. That’s neat, but how does it make sense?
How can perception stay stable if the neurons keep changing?
I’m sure manifolds will help us understand that too…
🧠🧪
3/3
Reposted by Ole Christian Sylte
Read the full paper ‘Dopaminergic action prediction errors serve as a value-free teaching signal’ in @natureportfolio.nature.com
@gatsbyucl.bsky.social @imperialcollegeldn.bsky.social
www.nature.com/articles/s41...
@gatsbyucl.bsky.social @imperialcollegeldn.bsky.social
www.nature.com/articles/s41...
Dopaminergic action prediction errors serve as a value-free teaching signal - Nature
Dopaminergic action prediction error signals are used by mice as a value-free teaching signal to reinforce stable sound–action associations in the tail of the striatum.
www.nature.com
May 14, 2025 at 3:09 PM
Read the full paper ‘Dopaminergic action prediction errors serve as a value-free teaching signal’ in @natureportfolio.nature.com
@gatsbyucl.bsky.social @imperialcollegeldn.bsky.social
www.nature.com/articles/s41...
@gatsbyucl.bsky.social @imperialcollegeldn.bsky.social
www.nature.com/articles/s41...
Really amazing!
April 17, 2025 at 2:26 PM
Really amazing!
Huge thanks to co-first author @antjekilias.bsky.social for impressive experimental work, and to Marlene Bartos and @jf-sauer.bsky.social for their supervision!
February 5, 2025 at 9:09 AM
Huge thanks to co-first author @antjekilias.bsky.social for impressive experimental work, and to Marlene Bartos and @jf-sauer.bsky.social for their supervision!
How does this help downstream neurons?
If drift is coordinated (preserving geometry), downstream neurons can compensate and read stable spatial maps. But if drift distorts the geometry, a downstream network fails under noise.
If drift is coordinated (preserving geometry), downstream neurons can compensate and read stable spatial maps. But if drift distorts the geometry, a downstream network fails under noise.
February 5, 2025 at 9:09 AM
How does this help downstream neurons?
If drift is coordinated (preserving geometry), downstream neurons can compensate and read stable spatial maps. But if drift distorts the geometry, a downstream network fails under noise.
If drift is coordinated (preserving geometry), downstream neurons can compensate and read stable spatial maps. But if drift distorts the geometry, a downstream network fails under noise.