|
In the present paper we estimate convergence rates of approximation for functions of bounded variation, for functions which are exponentially bounded and locally of bounded variation, for functions with derivatives of bounded vairation, and for functions which are exponentially bounded and with derivatives locally of bounded variation, using general integral operators with probability kernels. Let $L_n(f,x):=\int_D f( t) d_t m_n(x,t)$ be the sequence of positive linear operators where $D=[0,\infty)$ or $D=R$, and $m_n(x,\cdot)$ is a probability measure$D$ for each $x \in D$. The rate of convergence are determined by estimating $|L_n(f,x)-{1 \over 2} (f(x+)+f(x-))|$ and $|L_n(f,x)-f(x)|$ in terms of certain bounds. The generalesult is then applied to produce estimate for particular operators, such as Phillips operators, Baskakov operators, Mirakjan-Sz\''asz operators, Post-Widder operators, and Gauss-Weierstrass operators. Application to them indicates that our methods cannot be asymptotically improved.
|