Modeling Associative Memories Using Formal Concept Analysis

Associative memory is one of the primary functions of the human brain. In the literature, there are several neural networks based models that represent associative memory with the help of pattern associations. In this paper, we model the associative memory activity using Formal Concept Analysis (FCA), which is a standard technique for data and knowledge processing. In our proposal, patterns are associated with the help of object-attribute relations and the memory is represented using the formal concepts generated using FCA. We show that the extent and intent relations in the concepts help us to recall the patterns bi-directionally. Further, we model the pattern recall process for the given input even when the exact match is not found in the memory, using the concept hierarchies in the concept lattice.