In Section 3, we have studied the convergence performance of
least square learning algorithm in compressed neural networks
set. We have derived the upper bound of regression learning algorithms
by using the approximation property of neural networks
and covering number. In this section we discuss how our results
relate to other recent studies.