MonteCarloMeasurements.jl
MonteCarloMeasurements.jl copied to clipboard
Propagation of distributions by Monte-Carlo sampling: Real number types with uncertainty represented by samples.
It would be nice to be able to set the default number of particles for `..`, `±`, and `∓`.
- [x] `dot` can be expressed like gemm or gemv for one or two vectors of particles. - [ ] In-place `mul!` and friends - [ ] `axpy!` and friends...
Ref issue #22
Really enjoying this package, thank you for your work on it. For Soss.jl, here's a simple model: ```julia m = @model begin x ~ Normal(0,1) y ~ Normal(x^2, 1) |>...
Similar to #26 Instead of reducing particles to uncertainty by calling std, we can make use of the quantiles of the sample, at least if number of particles is large....
See implementation in BSON.jl https://github.com/MikeInnes/BSON.jl/blob/95dc736ec99dcd6bbf1abb59fdefece85a9bc3c7/src/extensions.jl#L104
Should be more robust than existing register functions. Also, consider using `@andreverse`
VonMises
Hej Fredrik! It would be great if this worked with a [Von Mises distribution](https://juliastats.org/Distributions.jl/stable/univariate/#Distributions.VonMises): ```julia julia> Particles(VonMises()) ERROR: MethodError: no method matching iterate(::VonMises{Float64}) Closest candidates are: iterate(::Union{LinRange, StepRangeLen}) @ Base...