- Autocorrelators were introduced as a theoretical concept by Donald Hebb in 1949 and were further analyzed by Amari in the 1970s. They store bipolar patterns by summing their outer products to create a connection matrix.
- The autocorrelator recalls patterns using a vector-matrix multiplication recall equation. For example, it can correctly recall the stored pattern A2 from a distorted input pattern A' by reducing the noise through the computation.
- Bidirectional associative memories (BAMs) are two-layer neural networks based on earlier autocorrelator models. They can store and recall pairs of patterns using a correlation matrix and recall equations to iteratively update input-output pairs until equilibrium is