Approximation to the $n$-th derivative using reproducing kernels.

433 Views Asked by At

For integrable functions defined on the real line, the normalized gaussian function approximates the convolution identity, Dirac Delta, in the sense that if $$g(t):=N_0e^{-x²}$$ (denoting the normalizing constant by $N_0$) and $$g_{a0}(x):=aN_0e^{-(ax)²}$$ then, for all $f$ integrable, $$\lim_{a\to\infty} f*g_{a0}=f$$ almost everywhere.

My first question is: Is this true?

My second question is: does the $n$-th derivative of the Gaussian function approximates the derivade operator by convolution? I mean writing $$g_{an}:=\frac{d^n}{dx^n} g_{a0}(x)$$ then $$\lim_{a\to\infty} f*g_{an}=f^{(n)}$$ and (third question) whether the $n$-th derivative operator includes or not the normalization constant $N_n$ for $g_{an}$.