Lumped mini-column associative knowledge graphs

This paper presents an extension of active neural associative knowledge graphs (ANAKG) to their mini-column form where each symbol is represented several times. We demonstrate that this new associative memory organization preserves all properties of ANAKG memories like storage of knowledge based on spatio-temporal input sequences, while increasing recall quality, memory capacity, and its resolution for short-term memory recall. The implemented model combines ANAKG associative neuron idea with the idea of a hierarchical temporal memory that uses a mini-column form of symbol representation. Performed tests confirm our claims of higher resolution and higher memory capacity of the new associative knowledge graph.