8 An Introduction to Many-Objective Evolutionary Optimization
293
Fig. 8.20 The red line
dominates the blue line
Fig. 8.21 Parallel plot of six
similar solutions (not
necessarily non-dominated)
8.4.3.3 Glyph Plot
A glyph plot is also known as radar, spider, or star plot. Glyph plots can be
considered similar to parallel plots, except that in glyph plots, each axis has different
directions as opposed to parallel lines. A glyph plot can be considered as a parallel
plot that has been bent and connected. Examples are shown in Figs. 8.22 and 8.23.
For minimization case, a solution is dominated when all points of the glyph are
farther away from the center compared to other glyphs.
Due to the similarities between glyph and parallel plots, the advantages and
disadvantages are also similar. Both plots are simple and easy to use, and they
can show domination relations, but they can be easily cluttered. For Pareto front
visualization, this technique also cannot distinguish the front shape.
8.5 Surrogate Model in Multi- and Many-Objective
Optimization
Let’s recall one of the challenges mentioned in Sect. 8.4.1: expensive evaluation.
When the evaluation of the objective functions has a high cost, the number of
evaluations must be kept low. In multi- and many-objective problems, this is even
293
Fig. 8.20 The red line
dominates the blue line
Fig. 8.21 Parallel plot of six
similar solutions (not
necessarily non-dominated)
8.4.3.3 Glyph Plot
A glyph plot is also known as radar, spider, or star plot. Glyph plots can be
considered similar to parallel plots, except that in glyph plots, each axis has different
directions as opposed to parallel lines. A glyph plot can be considered as a parallel
plot that has been bent and connected. Examples are shown in Figs. 8.22 and 8.23.
For minimization case, a solution is dominated when all points of the glyph are
farther away from the center compared to other glyphs.
Due to the similarities between glyph and parallel plots, the advantages and
disadvantages are also similar. Both plots are simple and easy to use, and they
can show domination relations, but they can be easily cluttered. For Pareto front
visualization, this technique also cannot distinguish the front shape.
8.5 Surrogate Model in Multi- and Many-Objective
Optimization
Let’s recall one of the challenges mentioned in Sect. 8.4.1: expensive evaluation.
When the evaluation of the objective functions has a high cost, the number of
evaluations must be kept low. In multi- and many-objective problems, this is even