Conversation
tim-at-topos
added
enhancement
jrb137
force-pushed
the
ecld
branch
2 times, most recently
from
46de908 to
7608522
Compare
tim-at-topos
force-pushed
the
ecld
branch
10 times, most recently
from
6331054 to
d1f675f
Compare
tim-at-topos
force-pushed
the
ecld
branch
3 times, most recently
from
783b5a6 to
7a6fad8
Compare
|
@jrb137 I think that |
tim-at-topos
force-pushed
the
ecld
branch
2 times, most recently
from
ff71847 to
fc32b58
Compare
broken test Ustr/Uuid -> Id removed TO-DOs usize -> bool
this test checks both the heights of the derivative towers, and that each morphism has been lifted to the correct height.
tim-at-topos
force-pushed
the
ecld
branch
3 times, most recently
from
dc7d32d to
ac964c7
Compare
Morphism generator types changes from "deg" to "Degree", etc.
Previously all derivatives were named the same thing, resulting in an error.
merge main
| \taxon{doctrine} | ||
| \import{macros} | ||
|
|
||
| \p{Degree-delay signed categories are categories graded in the monoid #{\mathbb{N} \times \mathbb{N} \times \mathbb{Z}_2}. The two copies of #{\mathbb{N}} label edges by the \em{differential degree} and \em{order of delay}. The order of the differential degree is meant to indicate which derivative of the target is affected (linearly) by the source. For example, an edge #{A \to B} with degree 2 indicates a contribution of the form #{\ddot{B} \mathrel{+}= kA} for some #{k \in \mathbb{R}}. The \em{order of delay} only has meaning with the assumption that the delays are modelled stochastically by Erlang distributions with the same time-scale parameter (see, for example, [here](https://en.wikipedia.org/wiki/Erlang_distribution)). The labelling in #{\mathbb{Z}_2} indicates the "sign" of the influence, as described in \ref{dct-0002}.} |
There was a problem hiding this comment.
Perhaps worth explaining that the reason you're talking about the Erlang distribution is that it's a sum of i.i.d. Gammas, where the number of summands is the "order of delay" in the ECLD.
| @@ -0,0 +1,4 @@ | |||
| //! Utilities for analyses of ECLDs | |||
There was a problem hiding this comment.
Are you planning to add more submodules here? If not, I'd suggest removing the extra directory level and having just stdlib/analyses/ecld.rs.
| domains.insert(name("f"), 0); | ||
| domains.insert(name("g"), 2); | ||
| domains.insert(name("l"), 0); | ||
| domains |
|
@tim-at-topos, this PR has become pretty stale, unfortunately. Is this something that you'd still like to push forward? One thing that gives me pause is that I still don't really understand atomization or why it needs to be so complicated, though I'm sure there's a reason. |
|
atomization is complicated for simple reasons, but really this PR should be preceded by an RFC! feel free to close this and I'll write an RFC |
|
Ok, thanks! Happy to read such a document if/when you find the time for it. |
3 participants
Implements a new logic of extended causal loop diagrams: causal loop diagrams labelled with differential degree and order of delay (both natural numbers).
This PR also adds a pre-processing function (
catlog/src/stdlib/analyses/ecld/atomisations.rs) for future analyses, which is currently unused but included simply because it's already written (along with a simple test). If preferred, this could be moved to a separate PR :-)To-do:
/math-docsdocumentation (@jrb137)/helpdocumentation (@tim-at-topos)