A Sparse Distributed Memory Capable of Handling Small Cues, SDMSCue

Abstract

In this work, we present Sparse Distributed Memory for Small Cues (SDMSCue), a new variant of Sparse Distributed Memory (SDM) that is capable of handling small cues. SDM is a content-addressable memory technique that relies on similar memory items tending to be clustered together in the same region or subspace of the semantic space. SDM has been used before as associative memory or control structure for software agents. In this context, small cues refer to input cues that are presented to SDM for reading associations; but have many missing parts or fields from them. The original SDM failed to handle such a problem. Hence, our work with SDMSCue comes to overcome this pitfall. The main idea in our work; is the projection of the semantic space on a smaller subspace; that is selected based on the input cue pattern, to allow for read/write using an input cue that is missing a large portion. The test results show that SDMSCue is capable of recovering and recalling information from memory using an arbitrary small part of that information; when the original SDM would fail. SDMSCue is augmented with the use of genetic algorithms for memory allocation and initialization. We think that the introduction of SDMSCue opens the door to more research areas and practical uses for SDM in general.