search

leftaroundabout / uncertainly-haskell

Uncertainty propagation of values such as physical measurements or approximate computations
GNU General Public License v3.0
5 stars 0 forks source link

Question about status of project #1

Open mvcisback opened 9 years ago

mvcisback commented 9 years ago

Hey, I was wondering what the status of this project is/if there are any plans to eventually publish on hackage?

I'm trying to move my data analysis tool chain to Haskell and good error propagation library is one of the big missing features (currently I've hacked something together based on http://rosettacode.org/wiki/Numeric_error_propagation#Haskell).

leftaroundabout commented 9 years ago

I have insofar abandoned this project as the plain representation – uncertainties with the same type as expectation value – has turned out not really suitable for the things I wanted to do. To express uncertainties in the multidimensional case, you can't really assume the deviations in the components to be independent. The appropriate type to allow for such covariances is a Hermitian operator, a density matrix. Implementing that efficiently isn't quite so simple, though it can be done in a big class of vector spaces. In fact, affine spaces or, even more generally, manifolds. That's the project I've primarily worked on over the last months; it contains types that generalise the notion of uncertainty-specifiers, but not yet with a stable interface.

Also, the manifolds package is really quite a big dependecy, since it uses hmatrix for the computations. So... I suppose it would be desirable to also have a simpler package for uncertainties. However, I don't have spare time to publish this (with a properly thought-through interface), not until the manifolds version is up and running at least.

But I certainly don't mind if you take any of my code and make a hackage package of it.

mvcisback commented 9 years ago

@leftaroundabout ah, ok. I'll have a quick look through the manifolds uncertainty specifiers to see if that fits my needs.