lscg-solver

A linear least square solver using the Conjugate Gradient method.

• Solve sparse linear least square problems.
• Solve sparse constrained linear least square problems using a lagrange method.
• Decompose dense matrices with full pivot LU to obtain solution and kernel (null space).
• Uses the Eigen library for sparse matrix computation and the conjugate gradient method.
• Compiled into WebAssembly for performance.

Install

npm install lscg-solver

Usage

You may require the solver as a node(commonjs) module:

let LSCGSolver = require("lscg-solver");

You may also use the solver directly as a UMD module:

<script src="node_modules/lscg-solver/dist/lscg-solver-umd.js"></script>

Once you have the solver imported, call the initialize method before using any of it:

LSCGSolver.initialize().then(() => {
    // Code that use the solver
});

Documentation

• See index.d.ts for a full declaration of the methods available.
• See test for examples on how to use the solver.

Development

To build the solver, you'll need to first install Emscripten. Make sure the emcc command is available in your PATH.

The build process was tested on macOS 10.13.4 with Emscripten 1.37.35.

Once the dependencies has been installed, simply clone the repository and run:

# Get the submodules (Eigen)
git submodule init
git submodule update

# Install required node modules
npm install

# Build
npm run build

# Test
npm run test

License

MIT License

lscg-solver uses the Eigen library, which can be obtained from http://eigen.tuxfamily.org/index.php?title=Main_Page.