Genomic and proteomic sequence recognition using a connectionist inference model.

In this paper a proposal for implementing a connectionist associative memory model (CAMM) based on a novel approach for recognising sequences is presented. The objective of the CAMM is to satisfy medium-high capacity and the retrieval of an arbitrary number of multiple associative memories that satisfy the stimulus input. The architecture is constructed on-the-fly and is dependent on the information in the training set. The model is composed of two stages; StageI and StageII. StageI is concerned with the development of a state space graph representing the training set and embedding that graph in a connectionist model. During retrieval a graph is produced that represents the candidate solutions; some spurious memories may infiltrate the solution space which is removed in StageII using conventional techniques.