Something I have not seen yet but hope exists and I just missed it: some notion of how much structure a given type of data can hold. For example, it seems obvious that in a comparison of math versus english,math would have much more structure available. Or maybe, since in principal you can describe all of math in english, what I really mean is structure density, because math is basically nothing else and english is a lossier map to the same thing and more besides.
I feel like if we had this concept, we could milk some gains from making predictions or at least assumptions about how much structure we should expect in unstructured data of various kinds.
Something I have not seen yet but hope exists and I just missed it: some notion of how much structure a given type of data can hold. For example, it seems obvious that in a comparison of math versus english, math would have much more structure available. Or maybe, since in principal you can describe all of math in english, what I really mean is structure density, because math is basically nothing else and english is a lossier map to the same thing and more besides.
I feel like if we had this concept, we could milk some gains from making predictions or at least assumptions about how much structure we should expect in unstructured data of various kinds.