pymanopt 0.1

James Townsend,Niklas Koep,Sebastian Weichwald — Thu, 07 Apr 2016 14:44:27 -0000

Pymanopt is a Python toolbox for doing manifold optimization that computes gradients and Hessians automatically. It builds upon the Matlab toolbox Manopt but is otherwise independent of it. Pymanopt aims to lower the barriers for users wishing to use state of the art manifold optimization techniques, by relying on automatic differentiation for computing gradients and Hessians, saving users time and saving them from potential calculation and implementiation errors.

Pymanopt is modular and hence easy to use. All of the automatic differentiation is done behind the scenes, so that the amount of setup the user needs to do is minimal. Usually only the following steps are required:

Instantiation of a manifold to optimise over
Defininition of a cost function to minimise
Instantiation of a Pymanopt solver

See project website (https://pymanopt.github.io) for more details and a tutorial.

mloss.org pymanopt

pymanopt 0.1