LBFGS

This repository contains a C++20 port of the Limited-memory BFGS implementation from LBFGS-Lite, with a number of usability improvements:

• error handling using the outcome library
• a simplified API similar to the Eigen Levenberg-Marquardt module
• the ability to specify the scalar type (e.g. float or double)

Limitations

• no step bounds
• no progress report
• no cancellation

These will be addressed in the future.

Usage example

The user must define a functor which computes the function and its gradient. Please have a look at https://github.com/foolnotion/lbfgs/blob/main/test/source/lbfgs_test.cpp