Hello people! Our teacher will ask us for a ML pro...
# getting-started
o
Hello people! Our teacher will ask us for a ML project in our next lab class. For this he has suggested using the python language along with Anaconda. I want to ask is is there any good machine learning platform for Java/Kotlin which is stable now? I know there is Tensorflow but the docs for Java are non existant.
https://www.tensorflow.org/api_docs/java/reference/org/tensorflow/package-summary
Copy code
WARNING: The API is currently experimental and is not covered by TensorFlow API stability guarantees. See README.md for installation instructions.
d
I didn’t have much luck with using TF from Kotlin. I think if time-to-implement is a consideration you’re much better off using Python (which even has libraries that make using TF easier AFAIK). You may consider asking in the #datascience channel but I think you’ll get a similar response.
o
Ok thnx
r
This is a pretty vast library: https://deeplearning4j.org/ I've never used it, but I've heard good things about it and there seems to be a fair amount of documentation and tutorials.
g
Oooh, that's a good question and I'd love to hear what other are using too. To get as near a python experience using Kotlin, you'd first need to get the Kotlin Kernels for Jupyter Notebooks. https://github.com/twosigma/beakerx Next handling data. I'd recommend krangl: https://github.com/holgerbrandl/krangl Then you could dive into the machine learning stuff. Here's a repo I can recommend. https://github.com/thomasnield/kotlin-data-science-resources
m
I am very familiar with python and TF, however KMath for kotlin mau attempt to eventually fully port the java variant of TF to kotlin. However the primary thing to first do would be finish porting numpy and scipy which math dealing with extremely high level tensors easy and tensorflow in python has a very close correlation to python numpy equivalents. I would like to see it ported over however I think it will be a while yet till this is done.