Abstract:In this paper we point out a converse result of the celebrated Jensen inequality for differentiable convex mappings of several variables and apply it to counterpart well-known analytic inequalities. Applications to Shannon's and Renyi's entropy mappings are also given.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the reverse results of Jensen's inequality, especially for multivariate differentiable convex functions. Specifically, the author proposes a reverse inequality of Jensen's inequality and explores the applications of this reverse inequality in certain specific inequalities as well as in Shannon entropy and Rényi entropy mappings.
### Background of the Paper
Jensen's inequality is an inequality widely used in probability theory and statistics, which describes the properties of convex functions on weighted averages. Formally, for a convex function \( f \) defined on a vector space \( X \), if \( x_i\in X \) and \( p_i\geq0 \) (\( i = 1,\ldots,m \)), and \( P_m=\sum_{i = 1}^m p_i>0 \), then:
\[ f\left(\frac{1}{P_m}\sum_{i = 1}^m p_i x_i\right)\leq\frac{1}{P_m}\sum_{i = 1}^m p_i f(x_i) \]
### Main Contributions of the Paper
1. **Reverse Inequality**: The author proposes a reverse result of Jensen's inequality, which is applicable to multivariate differentiable convex functions. Specifically, if \( f:\mathbb{R}^n\rightarrow\mathbb{R} \) is a differentiable convex function, and there exist vectors \( \psi,\phi\in\mathbb{R}^n \) such that \( \psi\leq x_i\leq\phi \), and there exist \( m,M\in\mathbb{R}^n \) such that \( m\leq\frac{\partial f(x_i)}{\partial x}\leq M \), then for all \( p_i\geq0 \) (\( i = 1,\ldots,m \)) and \( P_m=\sum_{i = 1}^m p_i>0 \), we have:
\[
0\leq\frac{1}{P_m}\sum_{i = 1}^m p_i f(x_i)-f\left(\frac{1}{P_m}\sum_{i = 1}^m p_i x_i\right)\leq\frac{1}{4}\|\Phi - \phi\|\cdot\|M - m\|
\]
where \( \|\cdot\| \) represents the Euclidean norm.
2. **Applications**: The author applies this reverse inequality to some specific inequalities, such as the arithmetic - geometric - harmonic mean inequality, etc. In addition, the application in Shannon entropy and Rényi entropy mappings is also explored.
### Main Conclusions
- **Shannon Entropy**: Through the above - mentioned reverse inequality, the author deduces a reverse inequality of Shannon entropy, providing an estimate of the gap between Shannon entropy and the maximum possible value.
- **Rényi Entropy**: Similarly, the author also gives a reverse inequality of Rényi entropy, further enriching the understanding and application of Rényi entropy.
In conclusion, the main purpose of this paper is to provide a deeper understanding of the properties of convex functions by proposing the reverse results of Jensen's inequality and applying them to the entropy concepts in information theory. This not only expands the theoretical framework of Jensen's inequality but also provides new analytical tools for practical problems.