Mixing Autoencoder With Classifier: Conceptual Data Visualization

In this paper, a neural network that is able to form a low-dimensional topological hidden representation is explained. The neural network can be trained as an autoencoder, as a classifier or as a mixture of both and produces a different low-dimensional topological map for each. When it is trained as an autoencoder, the inherent topological structure of the data can be visualized, while when it is trained as a classifier, a topological structure that is further constrained by a given concept, for example, the labels of the data, can be formed. Here, the resulting visualization is not only structural but also conceptual. The proposed neural network significantly differs from many dimensional reduction models, primarily in its ability to execute both supervised and unsupervised dimensional reduction and its ability to visualize not only the structure of high-dimensional data but also the concept assigned to them at various levels of abstraction.