The algorithm projects the LC- MVB weight vector onto the signal subspace, thus the effect of the noise subspace on the weight vector is reduced.

Through the cyclic subspace regression (CSR) being applied to determine the weight of the RBF networks, the RBF-CSR model is produced.