Home › ExtensionsMathNet

Converts a CsrMatrix to and from Math.NET Numerics' sparse matrix as a copy of the stored values, so a caller already on Math.NET can hand a Lodestar vectorizer's output to it, or the other way round, without densifying.

dotnet add package Lodestar.Extensions.MathNet

using Lodestar.Abstractions;
using Lodestar.Extensions.MathNet;
using MathNet.Numerics.LinearAlgebra;
using MathNet.Numerics.LinearAlgebra.Double;

Matrix<double> dense = DenseMatrix.OfRowArrays(
    [0.0, 3.0, 0.0],
    [0.0, 0.0, 0.0],
    [4.0, 0.0, 5.0]);

CsrMatrix csr = MathNetInterop.ToCsrMatrix(dense);
int stored = csr.NonZeroCount;   // 3

A conversion: the values in equal the values out, which the tests assert both ways. docs/equivalence.md maps each Python call to its C# counterpart, with every deliberate divergence.

An interop package (decision 0003), built for net10.0 and netstandard2.0:

• Lodestar.Abstractions 0.2.0 or later
• MathNet.Numerics 5.0.0 or later

• Reference: extensions-mathnet/conversion
• Changelog
• Performance
• All packages