Intuitionistic Fuzzy Divergence

• Intuitionistic fuzzy divergence is a family of measures that quantifies differences between intuitionistic fuzzy sets using an extended Jensen–Shannon framework.
• The approach addresses inadequacies in previous methods by enforcing strict axioms like identity, symmetry, and the triangle inequality for reliable comparison.
• Its practical applications enhance uncertainty handling in decision-making and pattern recognition, offering robust information processing under fuzzy conditions.

Intuitionistic fuzzy divergence refers to a family of quantitative measures for distinguishing and comparing intuitionistic fuzzy sets (IFSs) and intuitionistic fuzzy values (IFVs), with a principal focus on metrics induced through the Jensen–Shannon divergence. These measures address critical discrimination issues and axiomatic deficiencies inherent in prior distance and similarity constructions for IFSs, thereby enhancing information processing under uncertainty, especially in decision-making and pattern recognition frameworks (Wu et al., 2022).

1. Formal Framework: Intuitionistic Fuzzy Sets and Values

Let $X$ be a universe of discourse. An intuitionistic fuzzy set (IFS) $I$ over $X$ is specified as

$$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$

where $\mu_I: X \to [0,1]$ is the membership function, $\nu_I: X \to [0,1]$ is the non-membership function, and for every $x$, $\mu_I(x) + \nu_I(x) \le 1$. The indeterminacy degree (or hesitation) is $\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$.

An intuitionistic fuzzy value (IFV) $\alpha$ is an ordered pair $I$0 with $I$1, $I$2. The set of all IFVs is denoted by $I$3. Atanassov's partial order on $I$4 is given by

$I$5

The complement of $I$6 is $I$7.

2. Extension of Jensen–Shannon Divergence to Intuitionistic Fuzzy Domain

The classical Jensen–Shannon divergence (JSD) for probability distributions $I$8 and $I$9 is

$X$0

with $X$1.

For IFVs $X$2 and $X$3, Wu et al. define the IFV-JS divergence: $X$4 where

$X$5

and $X$6.

After algebraic manipulation,

$X$7

Normalization to $X$8 gives the strict intuitionistic fuzzy JS distance ("$X$9", Editor's term): $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$0 where $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$1 is the Endres–Schindelin function.

For IFSs $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$2 on finite $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$3 with IFVs $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$4, $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$5 and weights $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$6, $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$7: $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$8

3. Axiomatic Foundation of Strict Intuitionistic Fuzzy Measures

A strict intuitionistic fuzzy similarity measure (SIFSimM) $$I = \left\{ \frac{\langle\mu_I(x),\,\nu_I(x)\rangle}{x}\mid x \in X \right\},$$9 must satisfy:

• (S1) $\mu_I: X \to [0,1]$0
• (S2) $\mu_I: X \to [0,1]$1
• (S3) Symmetry: $\mu_I: X \to [0,1]$2
• (S4′) Strict distinctiveness: $\mu_I: X \to [0,1]$3 and $\mu_I: X \to [0,1]$4
• (S5) Extreme dissimilarity: $\mu_I: X \to [0,1]$5

A strict intuitionistic fuzzy distance measure (SIFDisM) $\mu_I: X \to [0,1]$6 must be a normalized metric whose dual similarity $\mu_I: X \to [0,1]$7 is a SIFSimM, with standard normalization, symmetry, identity, and triangle inequality. These axioms are extended pointwise for IFSs over $\mu_I: X \to [0,1]$8.

4. Properties and Superiority of $\mu_I: X \to [0,1]$9

Axiomatic Properties

$\nu_I: X \to [0,1]$0 fulfills SIFDisM requirements:

• Nonnegativity: $\nu_I: X \to [0,1]$1 via $\nu_I: X \to [0,1]$2.
• Identity: $\nu_I: X \to [0,1]$3.
• Symmetry: trivially inherited from $\nu_I: X \to [0,1]$4.
• Triangle inequality: holds via Endres–Schindelin lemma, using $\nu_I: X \to [0,1]$5 (applied to each coordinate and aggregated with Minkowski inequality).
• Strictness: For $\nu_I: X \to [0,1]$6, the partial derivatives of $\nu_I: X \to [0,1]$7 force $\nu_I: X \to [0,1]$8.
• Extreme dissimilarity: $\nu_I: X \to [0,1]$9, uniquely.

These properties are aggregated over $x$0 for full IFSs.

Discriminative Power

Existing measures, such as Xiao’s J-S distance, fail to distinguish IFVs with strict inclusion order (e.g., $x$1 vs. $x$2 within $x$3), leading to counter-intuitive assignments of distance $x$4 for non-endpoint values. In contrast, $x$5 assigns strictly increasing distances along the Atanassov partial order: $x$6 with $x$7, precisely reflecting $x$8.

5. Dual Similarity Measure and Induced Entropy

Dual Similarity

Define the strict JS-based similarity

$x$9

This similarity satisfies (S1)–(S5). For IFSs,

$\mu_I(x) + \nu_I(x) \le 1$0

Induced Intuitionistic Fuzzy Entropy

For IFV $\mu_I(x) + \nu_I(x) \le 1$1, the entropy induced by $\mu_I(x) + \nu_I(x) \le 1$2 is

$\mu_I(x) + \nu_I(x) \le 1$3

which satisfies Szmidt–Kacprzyk entropy postulates:

• (E1) $\mu_I(x) + \nu_I(x) \le 1$4 iff $\mu_I(x) + \nu_I(x) \le 1$5 is crisp.
• (E2) $\mu_I(x) + \nu_I(x) \le 1$6 iff $\mu_I(x) + \nu_I(x) \le 1$7.
• (E3) $\mu_I(x) + \nu_I(x) \le 1$8.
• (E4) Monotonicity under order inversion.

For IFS $\mu_I(x) + \nu_I(x) \le 1$9, $\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$0.

6. Comparative Analysis and Representative Example

Consider $\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$1, with

$\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$2

Atanassov order yields $\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$3 for $\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$4.

Measure	$\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$5	$\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$6	Discriminativity
Xiao’s $\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$7	$\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$8	Same as left	No discrimination
$\pi_I(x) = 1-\mu_I(x)-\nu_I(x)$9	$\alpha$0	$\alpha$1	Order-preserving strictness

For $\alpha$2, $\alpha$3, matching the inclusion order and affording genuinely strict discrimination—an improvement over previous J-S–based metrics (Wu et al., 2022).

7. Applications and Research Significance

Strict intuitionistic fuzzy divergence measures, especially $\alpha$4 and its similarity dual, serve as foundational tools for decision-making and pattern recognition under IFS frameworks. The strictness and full axiomatic compliance ensure these metrics can reliably distinguish between states with subtle relationships, resolving problematic degeneracies found in widely used prior measures. A plausible implication is robust improvement in information-theoretic analyses, aggregation, and clustering of highly uncertain data, wherever intuitionistic fuzzy systems are employed. These advances are theoretically grounded and validated through comparative analysis and canonical examples (Wu et al., 2022).