7.4
The following experiments illustrate the procedure for learning the
optimal 
from one exemplary instance and demonstrate the use
of the learned estimate to the recognition of other scenes and models.
The experiments are conducted as two parts: (1) Given one triplet 
as the exemplary instance, compute the optimal
for it; (2) then use the estimated 
to
recognize other scenes and models. The non-parametric learning algorithm
is used for (1) because the data size is too small to assume a
significant distribution. Convergence properties of the algorithm are
also shown.