STK: a Small (Matlab/Octave) Toolbox for Kriging
 STK_SAMPCRIT_EMMI_EVAL computes the EMMI criterion

 CALL: EMMI = stk_sampcrit_emmi_eval (ZP_MEAN, ZP_STD, ZI)

    computes the value EMMI of the Expected MaxiMin Improvement (EMMI) for a
    multi-objective minimization problem, with respect to the observed values
    ZI, assuming Gaussian predictive distributions with means ZP_MEAN and
    standard deviations ZP_STD.  The value of the criterion is computed
    approximately, using Monte Carlo simulations.  The input arguments must
    have the following sizes:

       * ZP_MEAN    M x P,
       * ZP_STD     M x P,
       * ZI         N x P,

    where M is the number of points where the EMMI must be computed, P the
    number of objective functions to be minimized, and N the current number of
    Pareto optimal solutions.  The output has size M x 1.

 CALL: EMMI = stk_sampcrit_emmi_eval (ZP_MEAN, ZP_STD, ZI, NSIMU)

    allows to change the number of simulations NSIMU used in the calculation of
    the criterion.

 NOTE

 1) The result depends only on the non-dominated rows of ZI.

 2) Multi-objective maximization problems, or mixed minimization/maximization
    problems, can be handled by changing the sign of the corresponding
    components of ZP_MEAN and ZI.

 3) Objective functions should be normalized for better performances.

 REFERENCES

   [1] Svenson, J., & Santner, T. (2016).  Multiobjective optimization of
       expensive-to-evaluate deterministic computer simulator models.
       Computational Statistics & Data Analysis, 94, 250-264,
       DOI: 10.1016/j.csda.2015.08.011.

 See also: stk_sampcrit_ehvi_eval