The training process makes the network to define its decision
boundaries.
Depend on Network Topology, training method and training parameters
(such as learning rate and epoche) such boundaries may be formed
in inclusive or exclusive paradigm.
Following Illustration shows a tight inclusive decision boundary
on a RBF network with 0.01 spread constant.
Using
a higher spread constant will enlarge the boundary..
Following figures show decision boundary on a MLP network with linear
transfer function at output layer.
add
to favorites
