QuasiMonteCarlo.jl includes LowDiscrepancySample which returns a Halton sequence for d > 1 or the 1-dimensional van der Corbut sequence.
The code can be updated to use the QMC package for HaltonSampling so that we can drop the dependency on HaltonSequences.
The package seems to be more active as well and it being under the SciML umbrella gives me confidence, that it is the better dependency for this.
It doesn't seem to skip numbers by default as it does for others, e.g. Sobol. Something we could potentially add in the future to reduce dependencies in higher dimensions.
We should make sure the results are exactly the same just in case there is something different I can't see at first glance.
QuasiMonteCarlo.jl includes
LowDiscrepancySample
which returns a Halton sequence ford > 1
or the 1-dimensional van der Corbut sequence.The code can be updated to use the QMC package for HaltonSampling so that we can drop the dependency on HaltonSequences.
The package seems to be more active as well and it being under the SciML umbrella gives me confidence, that it is the better dependency for this.
It doesn't seem to skip numbers by default as it does for others, e.g. Sobol. Something we could potentially add in the future to reduce dependencies in higher dimensions.
We should make sure the results are exactly the same just in case there is something different I can't see at first glance.