Genetic sparse distributed memory

Kanerva's 'sparse distributed memory' (SDM) is a type of self-organizing neural network which is able to extract a statistical summary from large volumes of data as it is being processed online. Genetic algorithms have been used to optimize the 'location address space' which corresponds to the mapping from the input layer to the hidden units in the neural network implementation of the sparse distributed memory. If treated as a global optimization problem, the genetic algorithm will attempt to optimize the sparse distributed memory so as to extract a single best statistical predictor. However, the real objective is to obtain not just a single global optimum, but to extract information about as many local optima as possible, since each local optimum in this particular definition of the search space represents a different and distinct data pattern that correlates with some output in which we may be interested. The implementation details of a genetic sparse distributed memory as well as modified algorithm designed to deal better with multiple data patterns are presented.<<ETX>>