3 ms·
> ... Differential Relational Calculus is an even older theory https://doi.org/10.1109/69.683760 https://doi.org/10.1109/69.683760 and is is a kind of derivativ
by zozbot234 5y ago
> ... Differential Relational Calculus is an even older theory https://doi.org/10.1109/69.683760 https://doi.org/10.1109/69.683760 and is is a kind of derivative on Relations using delta-sets. ...
That paper while prima facie relevant (so I appreciate that you mentioned it!), is about solving the problem of integrity maintenance, that is the only sense in which "differential" or "differentially" is used there, and then only as a matter of informal convenience and not as actively proposing any new terminology.
It does not define any true "differential" of a general computation as Differentiable Programming does in its own context, but only shows how to dynamize the checking of integrity constraints. So, I still maintain that uncritical adoption of such informal terminology, when clearly the proper alternative already exists in the literature, can only bring confusion and should be avoided.
- maweki 5y agoI'm sorry, but differential describes pretty good what's happening here. Given a Query Q and a database instance I, we calculate a Result R Q(I) = R And indeed we derive some function D for the differential and are given a set of changes C where D_Q(I)(C) = R' with Q(I+C)=R+R' This D is clearly a differential function. And if you look at the wikipedia entry for dynamic programming https://en.wikipedia.org/wiki/Dynamic_programming https://en.wikipedia.org/wiki/Dynamic_programming and indeed the important sentence "there is a relation between the value of the larger problem and the values of the sub-problems." In this case here there is no splitting into sub-problems going on. We only have a function (the Query) and we calculate how the result changes when the input changes. If that's not a differential, I don't know what is.