In this presentation, we present an improved set-membership partial-update
affine projection (I-SM-PUAP) algorithm, aiming at
accelerating the convergence, and decreasing the update rates
and the computational complexity of the set-membership
partial-update affine projection (SM-PUAP) algorithm. To
meet these targets, we constrain the weight vector perturbation
to be bounded by a hypersphere instead of the threshold
hyperplanes as in the standard algorithm. We use the distance
between the present weight vector and the expected update
in the standard set-membership affine projection (SM-AP)
algorithm to construct the hypersphere. With this strategy,
the new algorithm shows better behavior in the early iterations.
Simulation results verify the excellent performance of
the proposed algorithm related to the convergence rate and
the required number of updates.