topographical internal layers are good for hierarchical neural networks.

What is it about?

Mathematical derivations, supported by empirical experiments , showing that topological representations can be formed as the internal representations for layered network.

Featured Image

Why is it important?

As opposed to many deep models where the formation process of the internal representations is not always clear, in this paper, we propose a new neural network model in which the formation of the internal layers can be mathematically explained. It gives clearer understanding on the learning process of layered neural networks.

Perspectives