Introduction to Feed-Forward Hypercube Neural Networks (HNN)

摘要

One of the features that is lacking in the structure of a traditional Feed Forward multi-layer Artificial Neural Network (FFANN) is leapfrogging connectivity between the layers of neurons. The idea of Hypercube Neural Networks (HNN) is to model more realistically the practical features such as leapfrogging feed forward connections, computational efficiency, and its connection abstraction (abstract idea represented in a connectionist network) of multi-layer neural networks. HNN can be visualized as planes of FFANNs cascading together by Hyper links. Theoretical analysis of the proposed HNN is presented, followed by an implementation of its capability to demonstrate a real world application. The developed structure, theory and application of HNN contribute towards the broadening of the field of traditional FFANN's.