The capacity of the semi-orthogonally associative memories model

This paper discusses the capacity of the SAM model and demonstrates that, in this model, there exists a paralysis index such that the recalling outputs converge to the desired pattern on initial inputs when the initial similar probability is larger than this index, or is not true. For any given neurons' number N, this index is a function of the characteristic parameter. The authors show how to determine the optimum characteristic parameter n of this model. The memory capacity of this model is N/2 ln ln N, and the capacity for storing information at each synaptic connection is larger than 1/2/spl pi/ bits.