Classifier with Hierarchical Topographical Maps as Internal Representation

TL;DR

This paper introduces a hierarchical topographical map-based classifier that integrates top-down and bottom-up learning, enhancing deep representational learning with biologically inspired internal maps influenced by data labels.

Contribution

It presents a novel classifier architecture combining hierarchical topographical maps with deep learning, advancing previous models in complexity and biological relevance.

Findings

Enhanced internal representations in deep learning models

Effective integration of top-down and bottom-up learning processes

Improved performance in challenging classification tasks

Abstract

In this study we want to connect our previously proposed context-relevant topographical maps with the deep learning community. Our architecture is a classifier with hidden layers that are hierarchical two-dimensional topographical maps. These maps differ from the conventional self-organizing maps in that their organizations are influenced by the context of the data labels in a top-down manner. In this way bottom-up and top-down learning are combined in a biologically relevant representational learning setting. Compared to our previous work, we are here specifically elaborating the model in a more challenging setting compared to our previous experiments and to advance more hidden representation layers to bring our discussions into the context of deep representational learning.