Project Description and Background The Markowitz approach has already been tested at XXX to solve portfolio allocation problems (including Bayesian methods for finding parameters). The main problem encountered is that the classical estimation techniques do not produce dependable variance/covariance matrices and returns. We would like to test the so-called robust optimization solution techniques to see if they provide better results. The second part of the project would be to solve large-scale CVaR (Conditional VaR) models through the development of specific algorithms. In particular, tests should be conducted to see whether optimization algorithms for convex (non differentiable, with the idea to reduce the size of the original models) problems can improve the implementation of the CVaR approach. Another alternative could consist in implementing the Rockafellar & Uryasev approach with a freeware large scale optimization code. What is Involved? Development of interfaces and/or algorithms based on robust optimization as well as non differentiable optimization to solve financial risk models. Skills Needed Good programming skills (windows environment). Willigness to learn the MATLAB programming language. Reading TR-2002-03. Goldfarb, D., and Iyengar, G. "Robust Portfolio Selection Problems" (2002). Rockafellar, R.T. and S. Uryasev. Optimization of Conditional Value-At-Risk. The Journal of Risk, Vol. 2, No. 3, 2000, 21-41.