Orthonormalized associative memories

An associative memory model with an absolute formal and simple base is presented. This model works in the following way. When a question with a new pattern is addressed to the system, the processing element or elements representing the closest learning patterns are activated. These processing elements belong to an intermediate layer. Once this layer is activated, the associated response is obtained but the linear dependences among the learning output patterns do not introduce any distortion. The learning process is very fast. The recalling process happens in only one forward step. As an example, an application to closeness between region segment problems is considered