Quadratic optimizier

Developers
61

Since this is more or less a tehnical issue, please open an issue on my local Math fork: GitHub - rok-cesnovar/math: The Stan Math Library is a C++ template library for automatic differentiation of any order using forward, reverse, and mixed modes. It includes a range of built-in functions for probabilistic modeling, linear algebra, and equation solving. and we can go into details there.

We need to make the function templated and work with stan::math::var. That is all. It currently uses Eigen::MatrixXd everywhere and we need to make sure those accept vars.

We dont need to write a /rev version if it.

Its much eaasier for me to add this to a fork of Math and make stan call it from there as you propose. The current process is a bit clumsy for me, but obviously much easier to plug and play with off-the-shelf Rstan.

1 Like
62

Looks like you can’t create issues on a fork. Should I create the issue in the “misc” repo instead?

63

By default its turned off yes. Open now.

64

I thought you may be interested in this paper on a quadratic solver implemented on neural networks in pytorch https://arxiv.org/abs/1703.00443

1 Like