POEM Software

Developers

POEM is developed at the Applied Algebra group at at the Institute for Analysis and Algebra at Technische Universität Braunschweig and by associated members. Its developtment started at TU Berlin in July 2017.

Current Team

For planed new content, and content in development see the Upcoming Features.

Past Team Members

Henning Seidler

Version History

We give a brief overview about the current and the past versions of POEM and the main changes in the versions. For further details see the documentation coming with the software.

Latest Version

Version 0.2.1.0(a): Authors: H. Seidler and T. de Wolff; Release date: 07/04/2019.
Main Contributions: MOSEK 9 is added as a solver for SONC and SAGE decompositions. The setup is refined with additional options. Some additional small functionalities are added, and a series of internal adjustment and bug fixes is carried out.

Past Versions

Version 0.2.0.1(a): Authors: H. Seidler and T. de Wolff; Release date: 02/14/2019.
Contributions: A series of bugs fixed concerning imports in the setup.py file, coefficients being read from strings or the database, and the detect_infinity function in polynomial.py.
Version 0.2.0.0(a): Authors: H. Seidler and T. de Wolff; Release date: 01/30/2019.
Main Contributions: A major update introducing two main features: First, solving unconstrained polynomial optimization problems via SAGE decompositions. Second computing lower bounds and decompositions in exact arithmetic via SONC and SAGE for polynomials with rational coefficients. Also, the computation of local minimizers via gradient methods to obtain upper bounds is added and some internal structural adjustments are carried out.
Version 0.1.1.1(a): Authors: H. Seidler and T. de Wolff; Release date: 11/01/2019.
Contributions: Small adjustments, which became necessary due to new versions of cvxpy and sympy.
Version 0.1.1.0(a): Authors: H. Seidler and T. de Wolff; Release date: 07/24/2018.
Main Contributions: A small update featuring in particular an improved performance of our SONC algorithm due to a reduction of the number of variables in the reduction of the corresponding geometric program. The updated also includes a sparse output for coefficients of solutions and a bug fix regarding obtaining SONC decompositions.
Version 0.1.0.0(a): Authors: H. Seidler and T. de Wolff; Release date: 02/12/2018.
Main Contributions: Original release of the software allowing to solve unconstrained polynomial optimization problems via SONC and SOS certificates and providing an output of the decompositions via a python interface. Computions are carried out in Python or Matlab using cvxpy / cvx with various solvers. Remote execution of SOStools and Gloptipoly is possible, too.