[FRIAM] better simulating actual FriAM

[uǝlƃ ↙↙↙]

Excellent! Thanks. It's not clear to me why I get so confused. Every time I think about uncertainty and information, I have to Google concepts like entropy and negentropy and re-orient myself. I suppose I just don't do enough hands-on work with it to develop a tacit memory.

On 7/20/20 5:18 PM, David Eric Smith wrote:
> But if one did want to keep track of signs, I think in several sentences below where Glen is talking about the presence of limitations’ reducing the allowed variability in some distribution, we could say we use one or another _entropy_ measure to quantify the reduction in likely variability.  To the extent that one tries to characterize _information_ as Shannon did — a measure of how much ambiguity in a sample is reduced by having some bit of knowledge that rules out variations — then the reductions in entropy of the constrained ensemble relative to its prior would be called a gain of information in moving to the posterior from the prior.  So without worrying about the zero-point for either of these measures, or their resulting absolute signs, in many settings one would talk of the change of information’s being positive when the change of the corresponding entropy is negative.

-- 
↙↙↙ uǝlƃ