Portfolio optimization based on divergence measures

A new portfolio selection framework is introduced where the investor seeks the allocation that is as close as possible to his "ideal" portfolio. To build such a portfolio selection framework, the f-divergence measure from information theory is used. There are many advantages to using the f-divergence measure. First, the allocation is made such that it is in agreement with the historical data set. Second, the divergence measure is a convex function, which enables the use of fast optimization algorithms. Third, the objective value of the minimum portfolio divergence measure provides an indication distance from the ideal portfolio. A statistical test can therefore be constructed from the value of the objective function. Fourth, with adequate choices of both the target distribution and the divergence measure, the objective function of the f-portfolios reduces to the expected utility function.