Further results on relative, divergence measures based on extropy and their applications
Saranya P. ∗ , S.M.Sunoj · 2025
Read Further results on relative, divergence measures based on extropy and their applications by Saranya P. ∗ , S.M.Sunoj (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-07123.txt.
Introduction
Saranya and Sunoj develop new ways to measure how much a set of outcomes spreads out and how unsure it is. Their measures build on extropy, a quantity tied to entropy. They define dynamic versions of relative extropy for two cases. One covers the life still left in an item that is still in use. The other covers the time that has already passed since an event. The work studies the properties of the new measures and how they link to older ones.
The authors also give a way to estimate relative extropy straight from data, with no fixed shape assumed. They test this estimator in computer simulations. They then apply the measures to real-life data sets to show practical use. The paper adds to the set of tools for comparing spread and doubt between two sets of outcomes over time.
Catalog record
| CREATOR | Saranya P. ∗ , S.M.Sunoj |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-07123
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Further results on relative, divergence measures based on extropy and their applications Saranya P. ∗ , S.M.Sunoj Abstract This study explores information measures based on extropy, introducing dynamic relative extropy measures for residual and past lifetimes, and investigating their various properties. Furthermore, the study analyzes the relationships between extropy-based divergence with dynamic relative extropy and other extropy measures. A nonparametric estimator for relative extropy is developed, and its performance is assessed through numerical simulation studies. The practical applicability of the relative extropy is demonstrated through some real-life data sets. 1 Introduction In scientific experiments, probability assessments are often compromised by limited or inaccurate data, affecting statistical estimation and inference. To address this, Kerridge, (1961) introduced the Kerridge measure, which quantifies the discrepancy between two probability functions. For two absolutely continuous random variables X and Y with CDFs F and G , and PDFs f and g , respectively, the Kerridge inaccuracy measure is defined as I(F,G)=-\int_{0}^{\infty}{f(x)\log g(x)dx}. This measure can also be expressed as the sum of Shannon entropy H(F) (see Shannon, (1948) ) and Kullback-Leibler (KL) divergence ( Kullback and Leibler, (1951) ) as, I(F,G)=H(F)+K(F,G), where H(F)=-\int_{0}^{\infty}{f(x)\log f(x)dx} is the Shannon entropy and KL divergence is denoted by K(F,G)=\int_{0}^{\infty}{f(x)\log\frac{f(x)}{g(x)}dx} . Note that the inaccuracy measure reduces to the corresponding uncertainty measure when the distributions are identical. More recently, extropy has emerged as a complementary dual to entropy, offering a novel perspective on uncertainty quantification. According to Lad et al., (2015) , the extropy of the random variable X is defined as J(X)=-\frac{1}{2}\int_{0}^{\infty}f^{2}(x)dx. (1.1) Lad et al., (2015) also defined the concept of relative divergence measure based on extropy, as d(f,g)=\frac{1}{2}\int_{0}^{\infty}\left({f(x)}-{g(x)}\right)^{2}dx. (1.2) Hashempour and Mohammadi, 2024b () introduced an extropy-based dynamic cumulative past inaccuracy measure and studied its characteristic properties. Also, Hashempour and Mohammadi, 2024a () provided a new measure of inaccuracy based on extropy for record statistics between distributions of the n th upper (lower) record value and parent random variable which is given in the form, \xi J(X,Y)=-\frac{1}{2}\int_{0}^{\infty}f(x)g(x)dx. (1.3) Toomaj et al., (2023) derived some properties and several theoretical merits of extropy, and dynamic versions of extropy. The dynamic extropy inaccuracy for residual lifetimes, X_{t}=(X-t|X>t) and Y_{t}=(Y-t|Y>t) , defined by Hashempour et al., (2024) as \xi J_{r}(X,Y,t)=-\frac{1}{2}\int_{t}^{\infty}\frac{f(x)}{\bar{F}(t)}\frac{g(x)}{\bar{G}(t)}dx, (1.4) equals to J_{t}(X) when f(x)=g(x) . Let X_{(t)}=(t-X|X\leq t) and Y_{(t)}=(t-Y|Y\leq t) be two past lifetime random variables, then dynamic past inaccuracy measure defined by Mohammadi et al., (2024) is given as \xi J_{p}(X,Y,t)=-\frac{1}{2}\int_{0}^{t}\frac{f(x)g(x)}{F(t)G(t)}dx. (1.5) The discrimination information based on extropy and inaccuracy between density functions f(x) and g(x) is defined as J(f|g)=\frac{1}{2}\int_{0}^{\infty}[f(x)-g(x)]f(x)dx=\frac{1}{2}E_{f}[f(X)-g(X)]. (1.6) For nonnegative random variables X and Y , the dynamic extropy divergence between two distributions f and g is defined due to Mohammadi et al., (2024) , given by J_{r}(f_{t}|g_{t})=\frac{1}{2}\int_{t}^{\infty}\left(\frac{f(x)}{\bar{F}(t)}-\frac{g(x)}{\bar{G}(t)}\right)\frac{f(x)}{\bar{F}(t)}dx, (1.7) and the dynamic past extropy divergence is defined by J_{p}(f_{t}|g_{t})=-\frac{1}{2}\int_{0}^{t}\left(\frac{f(x)}{F(t)}-\frac{g(x)}{G(t)}\right)\frac{f(x)}{F(t)}dx. (1.8) Saranya and Sunoj, (2024) proposed a new relative extropy measure based on the survival function, studied its properties, usefulness in testing uniformity, and its applications. Recently, Saranya and Sunoj, (2025) introduced a new cumulative extropy measure of inaccuracy and divergence, and obtained their various properties and applications in image classification and reliability analysis. Although the literature includes symmetric and asymmetric information measures based on entropy and extropy, each with distinct advantages and applications, the interconnections between these measures remain relatively unexplored. Investigating these relationships can provide deeper insight into their properties and lead to significant results in information theory and reliability. Accordingly, in the present study we focus on exploring extropy-based information measures. The paper is organized as follows. In Section 2, we discuss some properties of relative extropy and extropy divergence. We explore the dynamic relative extropy for residual lifetimes and derive some results and properties along with those of dynamic extropy divergences and dynamic extropy inaccuracy in Section 3. The past lifetime scenario of these measures is discussed in Section 4. Section 4 investigates the estimation and real data analysis of the relative extropy. 2 Relative extropy and extropy divergence In this section, we study some properties of relative extropy and extropy divergence. For two nonnegative continuous random variables X and Y with pdfs f and g respectively, using ( 1.1 ), ( 1.2 ) and ( 1.3 ), we have the relationship between relative extropy, extropy inaccuracy, and extropy as: d(f,g)=2\xi J(X,Y)-J(Y)-J(X). The relative extropy is always nonnegative and zero if f(x)=g(x) for every x . The relationship between relative extropy and extropy divergence is given by the following theorem. Theorem 2.1 . Let X and Y be two nonnegative random variables with density functions f and g respectively. Then the relative extropy of X and Y can be defined as the sum of extropy divergences as follows. d(f,g)=J(f|g)+J(g|f)=\frac{1}{2}(E_{f}[f(x)-g(x)]+E_{g}[g(x)-f(x)]), (2.1) where J(g|f) is the extropy divergence from g to f with respect to g given as follows: J(g|f)=\frac{1}{2}\int_{0}^{\infty}[g(x)-f(x)]g(x)dx. Proof. We can split ( 1.2 ) as follows: \begin{split}d(f,g)&=\frac{1}{2}\int_{0}^{\infty}\left({f(x)}-{g(x)}\right)^{2}dx\\ &=\frac{1}{2}\int_{0}^{\infty}f^{2}(x)-2f(x)g(x)+g^{2}(x)dx\\ &=\frac{1}{2}\left(\int_{0}^{\infty}({f(x)}-{g(x)})f(x)dx+\int_{0}^{\infty}({g(x)}-{f(x)})g(x)dx\right)\\ &=J(f|g)+J(g|f)\end{split} ∎ Remark 2.1 . J(f|g) and J(g|f) can be both nonnegative and nonpositive, however, their sum is always nonnegative. Kullback, (1959) studied the approximation of the Kullback-Leibler information of f(x) and f(x;\theta+\Delta\theta) . In the following theorem, we provide a theoretical approximation to relative extropy of two similar densities, that is, between f(x,\theta) and f(x,\theta+\Delta\theta) as follows: Theorem 2.2 . Let X be a nonnegative random variable with density function f with parameter \theta . Then, d(f(x,\theta),f(x,\theta+\Delta\theta))\approx\frac{\Delta\theta^{2}}{2}\int_{0}^{\infty}\left(\frac{d}{dx}f(x,\theta)\right)^{2}dx. (2.2) Proof. \begin{split}f(x,\theta+\Delta\theta)&=f(x,\theta)+f^{\prime}(x,\theta)\Delta\theta+f^{\prime\prime}(x,\theta)\frac{(\Delta\theta)^{2}}{2!}+...\\ f(x,\theta+\Delta\theta)-f(x,\theta)&\approx f^{\prime}(x,\theta)\Delta\theta+f^{\prime\prime}(x,\theta)\frac{(\Delta\theta)^{2}}{2!}\\ \end{split} which implies (f(x,\theta+\Delta\theta)-f(x,\theta))^{2}\approx(\Delta\theta f^{\prime}(x,\theta))^{2}. Then, \frac{1}{2}\int_{0}^{\infty}(f(x,\theta+\Delta\theta)-f(x,\theta))^{2}dx\approx\frac{(\Delta\theta)^{2}}{2}\int_{0}^{\infty}f^{\prime}(x,\theta)^{2}dx. Thus the result follows. ∎ Corollary 2.1 . Let X follows exponential with the mean 1/\lambda , ( 2.2 ) becomes d(f(x,\lambda),f(x,\lambda+\Delta\lambda))\approx\frac{(\Delta\lambda)^{2}}{4\lambda} It is evident from Figure 1 that the relative extropy value between two exponential densities with parameters \lambda and \lambda+\Delta\lambda increases widely with increase in the \Delta\lambda . Figure 1 : Comparison of approximation with the actual value of relative extropy for \lambda=2 . Next, we consider some ordering between two random variables X and Y . We say that, X is said to be smaller (larger) than or equal to Y in the (i) Extropy ordering, denoted by X\leq_{ex}(\geq_{ex})Y if J(X)\leq J(Y) . (ii) Extropy divergence ordering, denoted by X\leq_{ed}(\geq_{ed})Y , if J(f|g)\leq(\geq)J(g|f) . Theorem 2.3 . X<_{ex}(>_{ex})Y\iff X>_{ed}(<_{ed})Y . Proof. Since J(f|g)=\xi J(X,Y)-J(X) and J(g|f)=\xi J(X,Y)-J(Y) , we obtain J(f|g)-J(g|f)=J(Y)-J(X) which gives J(Y)>(<)J(X)\iff J(f|g){>(<)}J(g|f) . ∎ Corollary 2.2 . Two continuous nonnegative random variables X and Y satisfy the additive extropy model J(Y)=J(X)+c, if and only if they satisfy the additive extropy divergence model J(f|g)+c=J(g|f). It also implies d(f,g)=2J(f|g)+c . In the following theorem, we obtain the bounds of extropy divergence based on the orderings. Theorem 2.4 . The following are the conditions for which J(f|g) and J(g|f) are nonnegative. (i) X>_{ed}Y\iff Y>_{ex}X\implies J(f|g)>0 (ii) Y>_{ed}X\iff Y<_{ex}X\implies J(g|f)>0 Proof. Since relative extropy is always nonnegative, it follows that J(f|g)+J(g|f)>0 . Given that J(Y)>J(X) , we obtain J(f|g)>J(g|f) , which implies J(f|g)-J(g|f)>0 , Adding these inequalities results in 2J(f|g)>0 . The proof of (ii) follows in a similar way. ∎ 3 Relative, inaccuracy and divergence measures of extropy for residual lifetimes This section discusses the dynamic scenario of the information measures for residual lifetimes. We focus on deriving some interesting relationships between those information measures. The following is the definition of dynamic relative extropy for residual lifetimes. Definition