strong uniform convergence rates are proved to be optimal for kernel-type estimator and weighted estimator of conditional function in the fixed design and random design, under certain conditions.
对随机与固定设计的回归函数,分别利用核估计和非参数加权估计,在核函数及权函数满足一条件下,本文证明了估计一致强收敛于待估函数的速度可达到最优。