Neural Abstraction Pyramid Architecture - Hierarchical Neural Networks for Image Interpretation - page 96

4.4 Conclusions
The previous chapter introduced the Neural Abstraction Pyramid architecture and
presented some simple examples for its use. The examples highlighted different
features of the architecture, like its ability to implement local normalization, the
cooperation and competition of features on different hierarchical levels, the effects
of the update order, and the extraction of invariant features.
All these examples were designed manually. No learning was used so far. The
following two chapters discuss, how unsupervised and supervised learning tech-
niques can be applied in the Neural Abstraction Pyramid framework.