UOpt library is a solver for unconstrained optimization. It is built for the training of neural networks in deep learning and other machine learning application.

UOpt has interfaces for the languages C, C++, R, MATLAB, and Python.

Uopt can exploit second order derivative or Hessian information in Newton methods. There is also a matrix-free Newton method for large scale problems.

UOpt is available for Windows and Linux from the download page. The library is free to use for optimization of up to 200 variables. Larger numbers require a license. To obtain a quote, please contact us.