Advertisement
  • Loading metrics

PLoS Computational Biology Issue Image | Vol. 16(2) February 2020

PLoS Computational Biology Issue Image | Vol. 16(2) February 2020

PLOS
x

Nonlinear mixed selectivity supports reliable neural computation

The arrangement of stimulus representations in neural population response space determines how robust these representations are to noise, as their arrangement determines which noise perturbations (grey arrows) can be corrected by a decoder (green arrow) and which cannot (red arrow). Johnston et al. shows that nonlinear mixed selectivity for multiple stimulus features produces stimulus representations that are well separated in response space and, as a result, far more robust to noise than selectivity that encodes different stimulus features either separately or with only linear mixing – even when both encodings use the same amount of metabolic resources.

Image Credit: W. Jeffrey Johnston, The University of Chicago

thumbnail
Nonlinear mixed selectivity supports reliable neural computation

The arrangement of stimulus representations in neural population response space determines how robust these representations are to noise, as their arrangement determines which noise perturbations (grey arrows) can be corrected by a decoder (green arrow) and which cannot (red arrow). Johnston et al. shows that nonlinear mixed selectivity for multiple stimulus features produces stimulus representations that are well separated in response space and, as a result, far more robust to noise than selectivity that encodes different stimulus features either separately or with only linear mixing – even when both encodings use the same amount of metabolic resources.

Image Credit: W. Jeffrey Johnston, The University of Chicago

https://doi.org/10.1371/image.pcbi.v16.i02.g001