Re: How to represent theories?

Thanks for all the input.

Le ven. 1 nov. 2019 à 15:45, Amirouche Boubekki
<amirouche.boubekki@gmail.com> a écrit :
>
> I stumbled upon an interesting problem based on my work on vnstore
> (formerly fstore) that is how to represent several theories made by an
> algorithm in the context of a versioned branch-able database (like
> git).


> What do you think? How do you handle alternative theories in your work?

I think I figured that theories are not meant to be tracked in git
branches because git branches are meant at some point to be merged
into the main branch or discarded. That is, in the *general case*, the
relation between theories can not reliably modeled using a
directed-acyclic-graph. In a subset of the design that support
theories there is the notion of lineage, that a theory stem from
another theory. In the general case, a theory can cherry pick piece of
knowledge discovered in another theory that does not have a common
ancestor, hence it looks more like graph than a tree. Also, in the
git-like approach it is made difficult for a given theory to discover
other theories because they live in different spaces. An algorithm
must explicitly look into other vnstore branches to find some
knowledge.

I seems like vnstore branches must be a tool for the user, not the algorithm.

Received on Thursday, 21 November 2019 09:26:01 UTC