Alpha-Beta Associative Memories

The Lernmatrix, which is the first known model of associative memory, is an heteroassociative memory, but it can also act as a binary pattern classifier depending on the choice of the output patterns. However, this model suffers two great problems: saturation and imperfect recall of some of the associations, even in the fundamental set, depending on the associations. In this work, a modification to the original Lernmatrix recall phase algorithm is presented. This modification improves the recalling capacity of the original model. Experimental results show this improvement

Memories have been one of the most important models for pattern recognition, being the Alpha-Beta Associative Memories the best available model today. In this paper we propose the use of Binary Decision Diagrams to represent Alpha-Beta associative memories. By doing so, we will be able to merge two important and very active areas of contemporary scientific research, and improve on both models.

Bidirectional Associative Memories (BAM) based on Kosko's model are implemented through iterative algorithms and present stability problems. Also, these models along with other models based on different methods, have not been able to perfectly recall all trained patterns. In this paper we present an English-Spanish / Spanish-English translator based on a new BAM model denominated Alpha-Beta BAM, whose process is non iterative and does not require to find stable states. The translator recalls the whole set of learned patterns, even when the presented word is incomplete.