Unconstrained minimum mean-square error parameter estimation with Hopfield networks

D.W. Tank and J.J. Hopfield (1986) have shown that an interconnected set of neuron-like elements can perform constrained minimum mean-square error (MMSE) estimation, such that estimated parameters are either zero or one. It is shown that a Hopfield network can also be applied to unconstrained MMSE estimation, such that estimated parameters can be any real number. Since unconstrained MMSE estimation is one of the most important operations in signal processing, the discovery that Hopfield networks can be used for such problems substantially increases their applicability.<>