Fig. 2: Transformation of trial-by-trial input variability to output variability.
From: Role of interneuron subtypes in controlling trial-by-trial output variability in the neocortex
a A normal distribution of inputs across trials (x-axis) is transformed into a distorted distribution of outputs (y-axis). Different colors denote different distributions. b Output trial-by-trial variability (measured as standard deviation) for different mean and variance across-trial inputs in a one-dimensional case. c Transfer-function of the E-I network with weak recurrency. The input clouds (denoted by colors) were sampled from distinct trial-by-trial input distributions. d Output rate distribution of corresponding input clouds shown in c for the E-I network. e Output trial-by-trial variability as a function of mean input to Exc and Inh populations for a fixed covariance matrix. \({\sigma }_{{{{{{{{\rm{E/I}}}}}}}}}^{{{{{{{{\rm{in}}}}}}}}}\) is unitless and \({\sigma }_{{{{{{{{\rm{EI}}}}}}}}}^{{{{{{{{\rm{in}}}}}}}}}\) has unit Hz2. The red(blue) dot denotes the red(blue) distribution in c. f Output trial-by-trial variability as a function of covariance \({\sigma }_{{{{{{{{\rm{EI}}}}}}}}}^{{{{{{{{\rm{in}}}}}}}}}\) and balance \({\sigma }_{{{{{{{{\rm{E/I}}}}}}}}}^{{{{{{{{\rm{in}}}}}}}}}\) (slanted bars illustrate the orientation of sampling cloud as shown in c), for a fixed mean input to the two populations. The red(black) dot denotes the red(black) distribution in c.
