Complete machine learning library for Java/Scala

Question

Python is plenty of ML libraries (like the great scikit-learn). Is there any good for java/scala, containing many algos (regression, classification, clustering, cross-validation, feature processing), stable & maintained and able to deal with massive dataset?

I've just found Mahout, Breeze/Nak, and Weka, but they're not looking as great as Python ones.

Additionally, if there's no equivalent, how can I efficiently connect java code with Python?

Perhaps consider adding a little detail about what makes scikit-learn great compared to Mahout &c. — Scortchi - Reinstate Monica, Jun 13 '14 at 13:05
Lots of algorithm for classification, regression, feature selection, feature reduction, and so on. Additionally, it has a big community of contributors and support — boskaiolo, Jun 13 '14 at 14:15
Could you be a bit more specific about what you're trying to achieve? A common way to "connect" java and python is to use jython, http://www.jython.org/. — sandris, Jul 17 '14 at 07:56

score 16 · Answer 1 · edited Aug 12 '15 at 15:27

16

You may find helpful this extensive curated list of ML libraries, frameworks and software tools. In particular, it contains resources that you're looking for - ML lists for Java and for Scala.

edited Aug 12 '15 at 15:27

Joshua Taylor

105
4

answered Aug 28 '14 at 07:14

Aleksandr Blekh

7,867
2
27
93

8

i am the owner of that list -i would like to add that it is updated regularly - with pull requests coming in every week – Joseph Misiti Jan 16 '15 at 01:35
@JosephMisiti: Thank you for the update. It's good to know - keep it up. – Aleksandr Blekh Jan 16 '15 at 03:06

score 5 · Answer 2 · answered Aug 27 '14 at 18:22

Apache Spark and specifically its component MLlib looks like exactly what you are looking for. MLlib contains implementations for classification, regression, dimensionality reduction etc. You can program in Scala,Java and Python.

Its basically a very fast distributed computing framework that can be run in an Hadoop cluster. For development purposes, you can easily run it in standalone mode (without Hadoop) on your local machine too.

Check out the MLlib guide here : https://spark.apache.org/docs/latest/mllib-guide.html

score 1 · Answer 3 · answered Aug 28 '14 at 06:57

Hava a look at JavaML (http://java-ml.sourceforge.net/) and Encog (http://www.heatonresearch.com/encog). The latter focuses rather on Neural Networks than on many algorithms.

Also, weka might not have very friendly java API (because, first of all, it's a GUI application, not a library), but when you get used to it, you start appreciating how many things are implemented there.

I have used successfully all of them.

Complete machine learning library for Java/Scala

3 Answers3