The inflection-point conjecture for the α\alpha-norm of a parametrized probability vector

From papers

Let n3n\ge 3, let α(0,1)(1,+)\alpha\in(0,1)\cup(1,+\infty), and let p[0,1n]p\in[0,\frac{1}{n}]. Write \boldmathvn(p)\text{\boldmath }v_n(p) for the probability vector under consideration and H\boldmathvn(p)H_{\text{\scriptsize \boldmath }v_n}(p) for its Shannon entropy. Let χn(α)\chi_n(\alpha) denote a possible inflection-point entropy.

Inflection-point conjecture. There exists an inflection point χn(α)(H\boldmathvn(en),lnn)\chi_n(\alpha)\in\bigl(H_{\text{\scriptsize \boldmath }v_n}(\mathrm{e}^{-n}),\ln n\bigr) such that \boldmathvn(p)α\|\text{\boldmath }v_n(p)\|_\alpha is strictly concave as a function of H\boldmathvn(p)H_{\text{\scriptsize \boldmath }v_n}(p) on [0,χn(α)][0,\chi_n(\alpha)] and strictly convex on [χn(α),lnn][\chi_n(\alpha),\ln n]. Moreover, χn(α)\chi_n(\alpha) is strictly increasing for α(0,1)(1,)\alpha\in(0,1)\cup(1,\infty), and

limα0+χn(0)>H\boldmathvn(en),limα1χn(α)=ln2+lnn1,limα+χn(α)=lnn.\lim_{\alpha\to0^+}\chi_n(0)>H_{\text{\scriptsize \boldmath }v_n}(\mathrm{e}^{-n}),\qquad \lim_{\alpha\to1}\chi_n(\alpha)=\ln 2+\ln\sqrt{n-1},\qquad \lim_{\alpha\to+\infty}\chi_n(\alpha)=\ln n.

This is presented as a more accurate description of the preceding lemma's convexity and concavity behavior. The supplied text gives no evidence that the claim has been proved or disproved, so its status remains open.

Progress summary

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Sources & referencesView supporting material

Primary source

Yuta Sakai and Ken-ichi Iwata, “Relations Between Conditional Shannon Entropy and Expectation of _α-Norm”, arXiv:1602.04577 (2016).

Solutions 0

No solutions have been posted yet.