Braden Brinkman
@professorbrink.bsky.social
Professor Brink Professor Brink he makes you laugh he makes you think
Theoretical/Computational Neuroscientist
Assistant Professor, Stony Brook University, NY
Theoretical/Computational Neuroscientist
Assistant Professor, Stony Brook University, NY
I guess we still haven't explored them enough
October 27, 2025 at 7:13 PM
I guess we still haven't explored them enough
Congrats to Ayesha!
October 13, 2025 at 12:51 PM
Congrats to Ayesha!
My department is joint college of arts & sciences and school of medicine, so there are two links -- ok to apply to both if not sure which is the right fit.
September 4, 2025 at 1:33 PM
My department is joint college of arts & sciences and school of medicine, so there are two links -- ok to apply to both if not sure which is the right fit.
The results in the paper are based on a stochastic field theory formalism, following co-senior author Gabe Ocker's 2023 paper journals.aps.org/prx/abstract.... In ongoing work we are using this path integral formalism to estimate the rates of metastable transitions between active states!
Republished: Dynamics of Stochastic Integrate-and-Fire Networks
Neural dynamics are typically described by neural field theories derived long ago using simplified neuron models. A new framework incorporates biophysical nonlinearities into these theories.
journals.aps.org
June 2, 2025 at 3:13 PM
The results in the paper are based on a stochastic field theory formalism, following co-senior author Gabe Ocker's 2023 paper journals.aps.org/prx/abstract.... In ongoing work we are using this path integral formalism to estimate the rates of metastable transitions between active states!
There is lots more to do, like investigating the roles of synaptic heterogeneity, synaptic plasticity and learning, and more. But I hope this paper provides a solid foundation for the field to explore these directions in models commonly used in theoretical neuroscience.
7/n, n = 7.
7/n, n = 7.
April 4, 2025 at 5:48 PM
There is lots more to do, like investigating the roles of synaptic heterogeneity, synaptic plasticity and learning, and more. But I hope this paper provides a solid foundation for the field to explore these directions in models commonly used in theoretical neuroscience.
7/n, n = 7.
7/n, n = 7.
An effect like this inhibitory tuning could be a potential explanation for the fact that some exp't studies find mean-field behavior while others observe anomalous scaling. (Though there are other possible explanations as well).
6/n
6/n
April 4, 2025 at 5:48 PM
An effect like this inhibitory tuning could be a potential explanation for the fact that some exp't studies find mean-field behavior while others observe anomalous scaling. (Though there are other possible explanations as well).
6/n
6/n
To the order of approx I work to, the spectrum of the synaptic connections is the key. If the maximum eigenvalue is an outlier, mean-field scaling holds. In the nets I study inhibition can move this outlier to the bulk spectrum, and anomalous scaling may emerge.
5/n
5/n
April 4, 2025 at 5:48 PM
To the order of approx I work to, the spectrum of the synaptic connections is the key. If the maximum eigenvalue is an outlier, mean-field scaling holds. In the nets I study inhibition can move this outlier to the bulk spectrum, and anomalous scaling may emerge.
5/n
5/n
One of the main challenges in applying RG methods to neural populations is the fact that neurons are not arranged in crystalline lattices, for which RG works well
I relax this restriction by studying nets w/ homogeneous modes and random connections (& nets that can be reduced to these)
4/n
I relax this restriction by studying nets w/ homogeneous modes and random connections (& nets that can be reduced to these)
4/n
April 4, 2025 at 5:48 PM
One of the main challenges in applying RG methods to neural populations is the fact that neurons are not arranged in crystalline lattices, for which RG works well
I relax this restriction by studying nets w/ homogeneous modes and random connections (& nets that can be reduced to these)
4/n
I relax this restriction by studying nets w/ homogeneous modes and random connections (& nets that can be reduced to these)
4/n
So far, lots of exp't data has been interpreted as consistent with criticality, but theory has been limited to simulations or re-interpreting models from physics as coarse-grained models of neural activity.
This is the first (afaik) RG theory applied to a model of spiking neurons.
3/n
This is the first (afaik) RG theory applied to a model of spiking neurons.
3/n
April 4, 2025 at 5:48 PM
So far, lots of exp't data has been interpreted as consistent with criticality, but theory has been limited to simulations or re-interpreting models from physics as coarse-grained models of neural activity.
This is the first (afaik) RG theory applied to a model of spiking neurons.
3/n
This is the first (afaik) RG theory applied to a model of spiking neurons.
3/n
Context: proponents of the "critical brain hypothesis" argue that it benefits computation for a neural circuit to be tuned close to a critical point -- a boundary between different phases of activity.
Directed perc = silent <-> active states
Ising = async <-> low/high firing states
2/n
Directed perc = silent <-> active states
Ising = async <-> low/high firing states
2/n
April 4, 2025 at 5:48 PM
Context: proponents of the "critical brain hypothesis" argue that it benefits computation for a neural circuit to be tuned close to a critical point -- a boundary between different phases of activity.
Directed perc = silent <-> active states
Ising = async <-> low/high firing states
2/n
Directed perc = silent <-> active states
Ising = async <-> low/high firing states
2/n