ND4J

What it is

ND4J (N-Dimensional Arrays for Java) is a scientific computing library for the JVM, providing high-performance n-dimensional arrays, linear algebra operations, and GPU acceleration. It serves as the foundation for Deeplearning4j and other Java-based ML frameworks.

ND4J provides Nd4j class for creating and manipulating n-dimensional arrays (INDArray), supports element-wise operations, linear algebra, broadcasting, reductions, and integration with deep learning frameworks like Deeplearning4j.

Installation

<dependency>
  <groupId>org.nd4j</groupId>
  <artifactId>nd4j-native-platform</artifactId>
  <version>1.0.0-M2.1</version>
</dependency>

Getting started

The smallest useful thing you can do with it, and what each part means.

Creating an array

java

12345import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.api.ndarray.INDArray;

INDArray array = Nd4j.create(new double[]{1, 2, 3, 4});
System.out.println(array);

Creates a 1D n-dimensional array and prints its contents.

Basic arithmetic

java

1234INDArray a = Nd4j.create(new double[]{1,2,3});
INDArray b = Nd4j.create(new double[]{4,5,6});
INDArray c = a.add(b);
System.out.println(c);

Performs element-wise addition of two NDArrays.

Advanced usage

Where the library earns its place over a simpler alternative.

Matrix multiplication

java

1234INDArray mat1 = Nd4j.create(new double[][]{{1,2},{3,4}});
INDArray mat2 = Nd4j.create(new double[][]{{5,6},{7,8}});
INDArray result = mat1.mmul(mat2);
System.out.println(result);

Performs matrix multiplication using ND4J's `mmul` function.

Broadcasting operations

java

1234INDArray mat = Nd4j.create(new double[][]{{1,2,3},{4,5,6}});
INDArray addVec = Nd4j.create(new double[]{10,20,30});
INDArray result = mat.addRowVector(addVec);
System.out.println(result);

Demonstrates broadcasting a 1D vector across rows of a 2D matrix.

Statistical operations

java

INDArray arr = Nd4j.create(new double[]{1,2,3,4,5});
System.out.println(arr.meanNumber());
System.out.println(arr.stdNumber());

Computes mean and standard deviation of an array.

GPU acceleration

java

// Use ND4J with CUDA backend for GPU computations
// Ensure nd4j-cuda-platform dependency is included

ND4J can leverage GPU for faster computations on large matrices or deep learning tasks.

Errors and fixes

The failures you are most likely to hit, and what actually resolves them.

IllegalStateException: Occurs if ND4J backend is not properly initialized. Check native library dependencies.
DimensionMismatchException: Thrown when shapes of arrays do not match for operations. Verify array dimensions.
OutOfMemoryError: Reduce array sizes or enable off-heap memory for large datasets.

Best practices

Use vectorized operations instead of loops for performance.
Leverage broadcasting for operations on arrays of different shapes.
Prefer Nd4j native platform or CUDA backend for large datasets or GPU acceleration.
Release unused NDArrays if memory usage is high.
Combine ND4J with Deeplearning4j for deep learning workflows.

Background

Why it exists, and what it was reacting to.

ND4J was created to bring NumPy-like capabilities to Java. It allows Java developers to perform vectorized operations, matrix computations, and linear algebra efficiently, both on CPU and GPU. ND4J is commonly used for deep learning, data analysis, and scientific computing within the Java ecosystem.

Related libraries

Others in AI & Machine Learning for Java.

JavaCV / OpenCV Java

Deeplearning4j (DL4J)

Weka

MOA (Massive Online Analysis)