Case Study: University of Geneva achieves fast, scalable portfolio optimization with MathWorks MATLAB

University of Geneva Develops Advanced Portfolio Optimization Techniques

The University of Geneva needed an automated, data-driven heuristic to solve complex portfolio-optimization problems involving advanced risk measures (like value at risk) and practical constraints that break classical techniques. Working with MathWorks, researchers used MATLAB and toolboxes (Parallel Computing Toolbox, MATLAB Distributed Computing Server, Optimization Toolbox, Financial Toolbox, and Statistics and Machine Learning Toolbox) but faced challenges tuning heuristic parameters and long compute times for many trial runs.

Using MathWorks tools, the team implemented a threshold-accepting heuristic in MATLAB, automated threshold sequencing, visualized intermediate results, and distributed dozens of independent starting-point runs across a 32-processor cluster with Parallel Computing Toolbox and MATLAB Distributed Computing Server. The solution reduced runtimes—single runs take 3–7 minutes and running 32 starts in parallel cuts what could be an 11+ hour serial task down to minutes—improved algorithm development through visualization, and gave students hands-on experience while matching classical Markowitz benchmarks.

University of Geneva

Manfred Gilli

Professor of the Department of Econometrics