A generalized proportionate adaptive algorithm based on convex optimization

A general framework is proposed to derive proportionate adaptive algorithms for sparse system identification. The proposed algorithmic framework employs the convex optimization and covers many traditional proportionate algorithms. Meanwhile, based on this framework, some novel proportionate algorithms could be derived too. In the simulations, we compare the new derived proportionate algorithm with the traditional ones, and demonstrate that it could provide faster convergence rate and tracking performance for both white and colored input in sparse system identification.