Weighted AM-GM is usually stated as follows:

given the non-negative reals $a_1,a_2,\dots,a_n$ and $\omega_1,\omega_2,\dots,\omega_n\ge 0$ with $\omega_1+\omega_2+\dots+\omega_n=1$ we have:$$\omega_1 a_1+\omega_2 a_2+\dots+\omega_n a_n\ge a_1^{\omega_1} a_2^{\omega_2}\dots a_n^{\omega_n}$$

Now, I was reading Mildorf's introduction to inequalities, he applies weighted AM-GM as follows:$$x_1+\frac{x_2^2}{2}+\frac{x_3^3}{3}+\dots+\frac{x_n^n}{n}\ge (1+\frac{1}{2}+\frac{1}{3}+\dots+\frac{1}{n})\cdot\:\sqrt[1+\frac{1}{2}+\frac{1}{3}+\dots+\frac{1}{n}]{x_1x_2x_3...x_n}$$ How does he apply weighted AM-GM to obtain that?

  • $\begingroup$ Let $H_n:=\sum_{i=1}^n 1/i$. Is your right-hand side meant to be $(H_n^{H_n})(\prod_i x_i)^{1/2}$, or $H_n(\prod_i x^i)^{1/H_n}$? $\endgroup$
    – J.G.
    Dec 7 '18 at 14:44
  • $\begingroup$ @J.G. The second one, sorry for the unclear LaTeX, it's $H_n\cdot \sqrt[H_n]{x_1x_2x_3...x_n}$ i.e. the $H_n$-th root of the product $\endgroup$ Dec 7 '18 at 14:52

A proof written while discovering the solution:

Define $H_n:=\sum_{i=1}^n\frac{1}{i}$ so the desired result is $\sum_i\dfrac{x_i^i}{iH_n}\ge\prod_ix_i^{1/H_n}$. Choosing weights $\omega_i$, the right-hand side is the weighted GM of the $x_i^{1/(\omega_i H_n)}$. Choose $\omega_i=\frac{1}{iH_n}$, so $\sum_i\omega_i=1$; the weighted GM of the $x_i^{1/\omega_i H_n}$ is then $\sum_i\dfrac{x_i^i}{iH_n}$ as required.

A neater version of the proof that doesn't show the discovery strategy:

With $\omega_i=\frac{1}{iH_n}$ and $x_i^i$ instead of $x_i$ throughout, $\sum_i\dfrac{x_i^i}{iH_n}\ge\prod_ix_i^{1/H_n}$ immediately follows; then multiply by $H_n$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.