User:Like.liberation/sandbox

ND4J is a free, open-source extension of the Java programming language operating on the Java Virtual Machine—though it is compatible with both Scala and Clojure.[1] It includes the Scala wrapper ND4S.[2]

ND4J is a scientific computing library for linear algebra and matrix manipulation in a production environment, integrating with Hadoop and Spark to work with distributed GPUs. It supports n-dimensional arrays for JVM-based languages.

ND4J has primarily been developed by the group in San Francisco that built Deeplearning4j, led by Adam Gibson.[3] It was created under an Apache Software Foundation license.

Distributed ND4J's operations include distributed parallel versions. They can take place in a cluster and process massive amounts of data. Matrix manipulation occurs in parallel on GPUs or CPUs in the cloud, and can work within Spark or Hadoop clusters.

Benchmarks Benchmarks show that ND4J runs roughly twice as fast as Numpy on large matrices.[4]