Re:Averaging is a convenient fiction of neuroscience
by JumpCrisscross on Monday September 23, 2024 @08:53PM
(#41632079)
> Averages are very bad in bimodal distributions
They're bad with multimodal distributions, generally, as well as any random process governed by a distribution with no mean, e.g. Cauchy [1].
(Neuronal firing appears to be non-Gaussian, possibly lognormal [2], which does have a mean [3], but it isn't equal to the simple average.)
[1] https://en.wikipedia.org/wiki/Cauchy_distribution#Properties
And that includes issues of public policy, where going left sort of works, and going right sort of works, and going in the middle sucks for absolutely everyone.