Classification Tree Generation Constrained with Variable Weights - Foundations on Natural and Artificial Computation - page 278

Fig. 2. Comparison of the relative weights in the correct trees for the complete A group
and for the first five species of A , Tree A (5) and Tree A (8) respectively. The left panel
shows a cross-plot of the weights. Values are in percentage.
the norm, but are necessary if one needs to overcome by statistical averaging the
problem of different features having different impacts in phylogeny. In general, it
is hard to collect data on hundreds of different features in fossilized organisms.
Furthermore, it is possible, at least in some cases, to assign weights to features
on the basis of a known phylogenetic tree and then extend those metrics to in-
clude additional organisms for which data may be highly incomplete. This raises
the possibility of extending the scope of these computations to cases where, until
now, data was insucient to support a phylogenetic tree.
Fig. 3. Comparison of the relative weights in the correct trees for the B and A groups
( Tree B (9) and Tree A (8) respectively, top panels) and for the A group plus the two
first species of B , compared to the weights of the A group ( Tree A (8) + B (2) and
Tree A (8) , bottom panels). The left panels shows the cross-plots of the weights. Values
are in percentage.