Synchronized Conceptual Representations in Generative Learning

In this work we examined latent representations of image data with a collective of generative neural network models. A convolutional autoencoder with steep redundancy reduction was used to create low-dimensional latent representations of a dataset of geometrical shapes. Individual models trained in entirely unsupervised process with minimization of generative error were then exposed to a process of synchronization of symbolic concepts associated with characteristic density structures in the latent representations. It was demonstrated that conceptual representations with good decoupling of concepts can be produced with generative models of limited depth; and that a simple process can lead to synchronization of symbolic concepts between learning individuals and a possibility of communication with sharing semantic information about the observed environment. The results demonstrate the potential of latent conceptual frameworks emergent in unsupervised generative learning as a natural platform for abstract conceptualization and communication.