Considering Memory Networks in the LTM Structure of the Multilevel Darwinist Brain

Abstract

In this paper, we present an initial implementation of the concept of memory network within the long term memory (LTM) structure of the Multilevel Darwinist Brain (MDB). The purpose is to address the problem of determining what, out of all the knowledge stored in the LTM, is relevant for the current context that a robot finds itself in, without having to resort to complex and time consuming search processes. Additionally, this type of structure allows for a very elegant way of hierarchically relating previously learnt knowledge with solutions to new situations. A very simple example is presented in order to help to understand how the system operates and its possibilities.