Java Numerical Processing (JNP) is an open-source multithreaded numerical computing package for the JVM which aims to provide efficient operations on dense matrices and tensors while maintaining a robust object-oriented design. JNP is currently a work-in-progress and in active development.
- An N-Dimensional Array type for matrix and tensor operations
- Data structure manipulation operations: Array resizing and reshaping, concatenation, transposition (for matrices), and slicing
- Advanced indexing (similar to MATLAB logical indexing)
- Numerical Operations
- Element-wise operations on N-dimensional data with support for complex valued data
- Vector operations with support for complex valued vectors
- Matrix Operations
- Linear equation solvers (planned)
- Numerical function integration and derivatives (planned)
- Graphing and visualization (planned)
- a set of Java Swing components for plots and data visualization that can be easily integrated into GUI based applications (planned)
- Domain specific tools (planned)
- I would like to add some domain specific functionality for signals and image processing, machine learning, and other fields.
All current functionality has been documented in javadoc comments in the source. More robust documentation will be made available when the tool reaches a more complete state.
More in-depth demos and documentation will be made available when the tool reaches a more complete state. Here are some quick functionality demos:
Create a random row and column vector:
Numeric rowVector = Numeric.Rand( 1, 4 );
Numeric colVector = Numeric.Rand( 4, 1 );
System.out.println( rowVector );
System.out.println( colVector );Output:
1x4 Array <Numeric>
[ 0.3965 0.8773 0.2981 0.4134 ]
4x1 Array <Numeric>
[ 0.6238 ]
[ 0.9725 ]
[ 0.5674 ]
[ 0.9671 ]
Create a 2x4 matrix by concatenating the row vector with itself along the 0th (row) dimension.
Numeric matrix = rowVector.concat( 0, rowVector );
System.out.println( matrix );2x4 Array <Numeric>
[ 0.3965 0.8773 0.2981 0.4134 ]
[ 0.3965 0.8773 0.2981 0.4134 ]
Create a 3x4 matrix by concatenating the transpose of the column vector.
matrix = matrix.concat( 0, colVector.transpose() );
System.out.println( matrix );3x4 Array <Numeric>
[ 0.3965 0.8773 0.2981 0.4134 ]
[ 0.3965 0.8773 0.2981 0.4134 ]
[ 0.6238 0.9725 0.5674 0.9671 ]
Reshape into a 2x6 matrix:
matrix = matrix.reshape( 2, 6 );
System.out.println( matrix );2x6 Array <Numeric>
[ 0.3965 0.8773 0.2981 0.4134 0.3965 0.8773 ]
[ 0.2981 0.4134 0.6238 0.9725 0.5674 0.9671 ]
Expand the 1st (column) dimension.
matrix.resize( 2, 9 );
System.out.println( matrix );2x9 Array <Numeric>
[ 0.3965 0.8773 0.2981 0.4134 0.3965 0.8773 0.0000 0.0000 0.0000 ]
[ 0.2981 0.4134 0.6238 0.9725 0.5674 0.9671 0.0000 0.0000 0.0000 ]
Create a 10x10 matrix.
Numeric matrix = Numeric.Rand( 10, 10 );
System.out.println( matrix );10x10 Array <Numeric>
[ 0.8145 0.4199 0.0836 0.8485 0.8000 0.3177 0.4427 0.1207 0.7857 0.0603 ]
[ 0.2164 0.2011 0.6299 0.4194 0.2847 0.0637 0.8694 0.5954 0.3255 0.2942 ]
[ 0.4490 0.9887 0.7508 0.9403 0.5298 0.9896 0.1718 0.5403 0.6593 0.9540 ]
[ 0.5082 0.1641 0.4453 0.3699 0.2321 0.2732 0.0205 0.5847 0.3965 0.2197 ]
[ 0.0040 0.3668 0.0354 0.2405 0.3725 0.0029 0.2416 0.8775 0.6379 0.5948 ]
[ 0.7303 0.8547 0.9954 0.7135 0.9801 0.5769 0.6543 0.5918 0.1025 0.5834 ]
[ 0.3909 0.5934 0.4403 0.4798 0.4765 0.7816 0.5186 0.1060 0.5212 0.0319 ]
[ 0.3787 0.9505 0.4516 0.2774 0.5172 0.1168 0.7896 0.1634 0.6183 0.4643 ]
[ 0.6386 0.5915 0.0199 0.9028 0.2993 0.3840 0.3881 0.1995 0.2334 0.5687 ]
[ 0.3849 0.0759 0.1741 0.5927 0.2101 0.8162 0.5438 0.1092 0.7767 0.0869 ]
Get a 4x5 slice of the last 4 rows and last 5 columns.
Numeric slice = matrix.slice( new int[]{6,10}, new int[]{5,10} );
System.out.println( slice );4x5 Array <Numeric>
[ 0.7816 0.5186 0.1060 0.5212 0.0319 ]
[ 0.1168 0.7896 0.1634 0.6183 0.4643 ]
[ 0.3840 0.3881 0.1995 0.2334 0.5687 ]
[ 0.8162 0.5438 0.1092 0.7767 0.0869 ]
Create a mask for all the elements in the matrix >= 0.5.
Logical mask = Element.greq( slice, Numeric.Scalar( 0.5 ) );
System.out.println( mask );4x5 Array <Logical>
[ 1 1 0 1 0 ]
[ 0 1 0 1 0 ]
[ 0 0 0 0 1 ]
[ 1 1 0 1 0 ]
Get a vector with only elements >= 0.5.
Numeric elements = slice.get( mask );
System.out.println( elements );1x9 Array <Numeric>
[ 0.7816 0.5186 0.5212 0.7896 0.6183 0.5687 0.8162 0.5438 0.7767 ]
Create a 5x5 matrix.
Numeric matrix = Numeric.RandComplex( 5, 5 );
System.out.println( matrix );5x5 Array <Complex>
[ 0.9233 +0.5330i 0.7210 +0.4225i 0.5493 +0.2910i 0.3486 +0.1879i 0.1821 +0.9264i ]
[ 0.0829 +0.5660i 0.3594 +0.9800i 0.6073 +0.9082i 0.3132 +0.3951i 0.8343 +0.8581i ]
[ 0.9029 +0.1563i 0.8597 +0.5054i 0.0453 +0.3708i 0.1828 +0.1735i 0.9205 +0.7533i ]
[ 0.7738 +0.6834i 0.5341 +0.2503i 0.3845 +0.8570i 0.7930 +0.8476i 0.0922 +0.9642i ]
[ 0.1134 +0.4378i 0.4449 +0.7994i 0.7209 +0.9965i 0.7296 +0.0792i 0.2250 +0.2129i ]
Compute sinh(z) + cos(z) in parallel using 8 threads.
ElementWiseExectutor E = new ElementWiseExecutor( 8 );
matrix = E.add( E.sinh(matrix), E.cos(matrix) );
System.out.println( matrix );5x5 Array <Complex>
[ 1.6040 +0.2950i 1.5353 +0.2340i 1.4423 +0.1772i 1.3060 +0.1337i 1.5465 +0.6199i ]
[ 1.2305 +0.4887i 1.6273 +0.4821i 1.5808 +0.3458i 1.3204 +0.2790i 1.5456 +0.3189i ]
[ 1.6450 +0.1003i 1.5862 +0.2750i 1.1107 +0.3456i 1.1792 +0.1439i 1.5558 +0.3370i ]
[ 1.5506 +0.3145i 1.4302 +0.1551i 1.5467 +0.4502i 1.5508 +0.3192i 1.5482 +0.7219i ]
[ 1.1933 +0.3756i 1.5272 +0.4072i 1.5826 +0.2952i 1.5414 +0.0482i 1.2188 +0.1688i ]
Graph f(x) = cos(x) + x*sin(x) for x = [-50,50].
NumericExpression1D f = x -> Math.cos(x) + x*Math.sin(x);
Numeric X = Numeric.LinSpace( -50, 50, 1000 );
Numeric Y = E.evaluate( f, X );
JFrame frame = new JFrame();
frame.setSize(800, 600);
frame.setTitle("JNP Plot");
frame.setDefaultCloseOperation(JFrame.EXIT_ON_CLOSE);
frame.setLocationRelativeTo(null);
LinePlot plot = new LinePlot( X, Y, 800, 600 );
frame.add(plot);
frame.setVisible(true);