I'm considering developing a Java alternative to NumPy using the vector API, although I know it's still in the incubator stage. I'm curious if there are any real use cases for this kind of library. Alternatively, would it be more beneficial to call a Python program through JNI? I'm still learning, so any insights would be appreciated!
1 Answer
Instead of going through Python, you might want to explore directly calling C functions using FFM. That's essentially what Python does, and if you're already working in Java, it makes sense to skip that extra step.
Are you saying creating a Java equivalent of NumPy is a viable idea? The downside is that it wouldn't match NumPy's speed since it's backed by C. So, what's the point? Shouldn't we focus on a Java-native solution like JNum for enterprise use instead of routing everything through Python for machine learning or data analysis?