Variable step-size diffusion proportionate affine projection algorithm

摘要

A combine-then-adapt (CTA) diffusion proportionate affine projection algorithm (DP APA) is proposed for distributed estimation, which uses gain matrices in the CTA diffusion affine projection algorithm (DAPA) to proportionately adapt the weight vectors of agents in the network. Then, a variable step-size (VSS) is presented for the DPAPA to address the problem of tradeoff between fast convergence rate and low steady-state misalignment. The VSS is developed by using the method of shrinkage, which uses a combined h- and I2-norm optimization to obtain the noise-free a priori errors. Simulation results verify that the VSS-DPAPA can obtain good performance.