Shared memory nodes and discriminant's reduced storage in two dimensional modified Kanerva's sparse distributed memory model

P. Kanerva's (1988) SDM classic model has been modified using a random memory space, whose qualities in two dimensions allow a significant reduction in the memory requirements of the modified net. The memory is managed in a space sharing mode; we present techniques that allow this kind of memory management, and we show that the operation of this kind of Kanerva net in two and three dimensional configurations works better than the conventional one dimensional case. We also introduce a short term memory model that increases the modified net's recognition performance.