Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


A Toolbox for Distributed Convex Optimization


Niklaus Voellmy

Master Thesis, FS 11 (10069)

In the following thesis, a solver tool for distributed convex optimization is introduced. The tool allows for an intuitive problem formulation in Matlab. Thereafter, C-libraries can be compiled with the tool. Finally, the libraries can solve the optimization problem on a distributed hardware system. The tool is built up modularly. Hence, different communication methods, local solver software, and distributed optimization algorithms can be combined arbitrarily. The implemented distributed optimization algorithms include dual decomposition (DD), a parallel version of the method of multipliers (PMM) and the alternating direction method of multipliers (ADMM). The dual variables in the dual decomposition algorithm can be updated with the fast gradient method. The toolbox and the optimization algorithms were evaluated on several benchmark problems. All investigated problems could be formulated and solved with the toolbox. Concerning the algorithms, ADMM and PMM were superior to DD in terms of usability as they need fewever assumptions to converge. ADMM and PMM also converged faster than DD, even if the fast gradient method was used on the latter. iii


Type of Publication:

(12)Diploma/Master Thesis

C. Conte

No Files for download available.
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2011:IFA_4132
Permanent link