AV
Adrian Valente
Author with expertise in Neural Interface Technology
Achievements
Open Access Advocate
Cited Author
Key Stats
Upvotes received:
0
Publications:
1
(100% Open Access)
Cited by:
20
h-index:
5
/
i10-index:
5
Reputation
Biology
< 1%
Chemistry
< 1%
Economics
< 1%
Show more
How is this calculated?
Publications
150

The role of population structure in computations through neural dynamics

Alexis Dubreuil et al.Jul 4, 2020
Abstract Neural computations are currently investigated using two separate approaches: sorting neurons into functional populations, or examining the low-dimensional dynamics of collective activity. Whether and how these two aspects interact to shape computations is currently unclear. Using a novel approach to extract computational mechanisms from networks trained on neuroscience tasks, here we show that the dimensionality of the dynamics and cell-class structure play fundamentally complementary roles. While various tasks can be implemented by increasing the dimensionality in networks with fully random population structure, flexible input-output mappings instead required a non-random population structure that can be described in terms of multiple sub-populations. Our analyses revealed that such a population structure enabled flexible computations through a mechanism based on gain-controlled modulations that flexibly shape the dynamical landscape of collective dynamics. Our results lead to task-specific predictions for the structure of neural selectivity, inactivation experiments, and for the implication of different neurons in multi-tasking.