Geometric Back-Propagation in Morphological Neural Networks
This paper provides a definition of back-propagation through geometric correspondences for morphological neural networks. In addition, dilation layers are shown to learn probe geometry by erosion of layer inputs and outputs. A proof-of-principle is provided, in which predictions and convergence of morphological networks significantly outperform convolutional networks.
Email Address of Submitting Authorr.firstname.lastname@example.org
ORCID of Submitting Author0000-0001-5218-0558
Submitting Author's InstitutionUniversity of Amsterdam
Submitting Author's Country