Correction to "Linear and Logarithmic Capacities in Associative Neural Networks"

J. Ma, J. Wu, and Q. Cheng kindly pointed out that [1, Proof of Proposition 1(b)] contains an overcount. The statement of the proposition reads: Let n be the family of bases for with all basis elements constrained to be binary n-tuples; (i.e., E = fe1; e2; . . . ; eng 2 n if, and only if, e1, e2; . . . ; en 2 f 1;+1g are linearly independent [over the field of the reals]). Then, asymptotically as n ! 1, almost all vectors u 2 f 1;+1g have a representation of the form