ETH Zuerich - Homepage
Systems Optimization (SOP)

Wichtiger Hinweis:
Diese Website wird in älteren Versionen von Netscape ohne graphische Elemente dargestellt. Die Funktionalität der Website ist aber trotzdem gewährleistet. Wenn Sie diese Website regelmässig benutzen, empfehlen wir Ihnen, auf Ihrem Computer einen aktuellen Browser zu installieren. Weitere Informationen finden Sie auf
folgender Seite.

Important Note:
The content in this site is accessible to any browser or Internet device, however, some graphics will display correctly only in the newer versions of Netscape. To get the most out of our site we suggest you upgrade to the latest Netscape.
More information

ETH Zürich - D-ITET - TIK - Downloads & Materials - Supplementary Materials - Testproblems - Zdt3
print page
this webpage might no longer be updated more...
Untitled Document


Pareto Front:

with and where

Relevant Publications:
  • E. Zitzler, K. Deb, and L. Thiele. Comparison of Multiobjective Evolutionary Algorithms: Empirical Results. Evolutionary Computation, 8(2):173-195, 2000 (PDF) (bibtex)
Reference Point: Reference Point used: (11,11)

where (see Formulation)

Optimal Distributions:
(see "Maximum Hypervolume" for more plots)
of 5 points:
of 10 points:
of 20 points:
of 50 points:
Maximum Hypervolume:
µHV ValuesPlot

How to approximate the optimal distributions

  1. The x-values xi of the µ points are equally distributed between 0 and 0.851832865436414
  2. The first and last point are known to be extremal, hence only the remaining µ -2 points are optimized.
  3. For all points p, starting with point p = 2, the following steps are executed:
    • The x-value is decreased by stepsize. If this increases the hypervolume, the procedure continues with the next point.
    • If decreasing xp does not increase the hypervolume, then xp is increased by 2*stepsize. If this does not increase the hypervolume, xp decreased by stepsize (which means it has the value at the start of step 3).

  4. If step 3 did increase the hypervolume, step 3 is repeated as long as the hypervolume increases. Otherwise, step 5 is executed
  5. For all points p, xp is set to a random value between 0 and 0.851832865436414 until the hypervolume increases or for maximum 10 times. This allows points to jump to different sections of the Pareto front.
  6. If the hypervolume increased in step 5, the procedure starts all over again with step 3.
  7. If no jump in step 5 did increase the hypervolume, stepsize is scaled down by 1/2. If stepsize is smaller than a predefined value eps (10-16), the precedure returns the current distribution. Otherwise, the precedure restarts with step 3.
© 2018 Institut TIK, ETH Zürich | Imprint | Last updated: Wed, 21 Mar 2018 12:31 | Valid XHTML 1.0! Valid CSS! Valid XHTML 1.0