Exact data retrieval of associative memory based on cross talk formulation

Based on mathematical formulation of cross talk, a new associative memory architecture that completely eliminates cross talk due to the memorizing keys having odd number Hamming distance from the input key has been presented (Y. Kumagai et al., 1992). An architecture by utilizing further advanced properties of the formulated cross talk which are theoretically guaranteed to be able to reduce the remaining cross talk is proposed. Compared to the conventional architecture that needs monotonic increment of the degree of high-order correlation and the architecture proposed earlier, it is shown that the architecture proposed here with somewhat more high-order cross products than the previously proposed one substantially improves exact data retrieval ability