For NDArrays, is Viktor or MultiK recommended, or ...
# datascience
m
For NDArrays, is Viktor or MultiK recommended, or is there a different library I should take a look at?
p
It depends on your tasks. Viktor is developed for bioinformatics tasks, so it has a limited number of methods. However, it uses SIMD, which provides very good performance. Only float64 is supported. Multik is more broadly applicable. It supports all major types and complex numbers. There are transformation methods for ndarray and some standard methods from linear algebra. There is also the kmath library. It offers a different approach to the API and wraps many implementations, including Multik and Viktor.
m
I ended up writing my own (largely based on Viktor), since there didn't seem any good multiplatform library utilizing SIMD instructions: https://github.com/Martmists-GH/ndarray.simd