Independent Sets in Graphs

Updated 30 January 2026

Independent sets in graphs are vertex subsets with no adjacent pairs, where the independence number marks the largest such set.
They are central to combinatorics and extremal graph theory, with concepts like ker, diadem, and independence polynomials providing structural insights.
Algorithmic research targets NP-hard maximum independent set problems using dynamic programming, heuristic search, and spectral methods.

An independent set in a finite simple graph $G=(V,E)$ is a subset $I\subseteq V$ such that no two vertices in $I$ are adjacent. Independent sets play a central role in extremal graph theory, combinatorics, algebraic graph theory, statistical mechanics, and computational complexity. Their enumeration, structural theory, optimization, and algorithmic properties have motivated a rich array of foundational and modern research directions.

1. Definitions and Key Parameters

Given a graph $G=(V,E)$ , an independent set is any $I\subseteq V$ with $\forall u,v\in I$ , $\{u,v\}\notin E$ . The independence number $\alpha(G)$ is the maximum size of an independent set. The set of all independent sets is denoted $\mathcal I(G)$ , and their total number is $i(G) = |\mathcal I(G)|$ . For fixed $k$ , denote $i_k(G) = |\{I\in\mathcal I(G) : |I|=k\}|$ .

For $X\subseteq V$ , define the neighborhood $N(X) = \{v\in V\setminus X: \exists u\in X, \{u,v\}\in E\}$ , and the difference $d(X) = |X| - |N(X)|$ . The critical difference $\delta^*(G) = \max\{d(X): X\subseteq V\}$ . A set is critical if $d(X) = \delta^*(G)$ . The ker and diadem of $G$ are

$\mathrm{ker}(G) = \bigcap \{X: X\text{ critical}\}, \quad \mathrm{diadem}(G) = \bigcup \{X: X\text{ is critical and independent}\}.$

The ker is always an independent set and satisfies $\mathrm{ker}(G) \subseteq \mathrm{core}(G)$ , where $\mathrm{core}(G)$ is the intersection of all maximum independent sets. An independent set $S$ is inclusion-minimal with $d(S)>0$ if for all proper subsets $Y\subset S$ , $d(Y)\leq 0$ .

2. Structural Results and Extremal Bounds

The enumeration of independent sets is constrained by various global and local properties of the graph:

In $n$ -vertex graphs of minimum degree at least $d = o(n^{1/3})$ , the complete bipartite graph $K_{d,n-d}$ uniquely maximizes $i_k(G)$ for all $k \geq 3$ , and uniquely maximizes the distribution and expectation of the size of a uniformly random independent set ( $X(G)$ ) among all such graphs (Law et al., 2012).
For regular graphs, Sapozhenko's upper bound is

$i(G) \leq 2^{\left(1 + C\sqrt{\frac{\log d}{d}}\right)\frac{n}{2}},$

and the sharp bound due to Kahn–Zhao asserts

$i(G) \leq (2^{d+1}-1)^{n/(2d)},$

with equality for a disjoint union of $K_{d,d}$ components (Samotij, 2014).

For certain structured graph classes, precise enumeration is tractable, e.g.:

Graph Family	Algorithmic Complexity	Reference
Cocomparability Graphs	$O(n+m)$ via poset decomposition	(Dyer et al., 2018)
Strongly Orderable	$O(n^2)$ via strong ordering	(Heinrich et al., 2021)
Distance-Hereditary	Poly-time via DP on cograph trees	(Chang et al., 2012)
Grid Graphs	$O(n2^{3m})$ , state-matrix recursion	(Oh et al., 2016)

3. Algebraic and Geometric Perspectives

The monomial ideal of independent sets $I_G$ in the polynomial ring $T=K[s_1,\ldots,s_n;t_1,\ldots,t_n]$ is

$I_G = (m_S : S\subseteq V, S \text{ indep.}), \quad m_S = \left(\prod_{i\in S} s_i\right) \cdot \left(\prod_{i\not\in S} t_i\right).$

The minimal prime decomposition reflects the covering structure of $G$ : $I_G = \left( \bigcap_{i\in V} (s_i, t_i)\right) \cap \left(\bigcap_{\{i,j\}\in E} (t_i, t_j)\right).$ Cohen–Macaulayness holds iff $G=K_n$ . The Betti numbers are determined by the coefficients $s_k$ of the independence polynomial $I(G;x)$ as

$\beta_i(I_G) = \sum_{k=0}^{\alpha(G)} s_k\cdot \binom{k}{i}.$

(Olteanu, 2013)

4. Enumerative and Probabilistic Theory

Enumerating independent sets is closely linked to the independence polynomial $P_G(\lambda) = \sum_{S\in I(G)} \lambda^{|S|}$ . Advanced combinatorial and analytic methods yield asymptotic expansions for various graph families:

For regular bipartite expanders,

$i(G) = 2^{n/2+1}\exp\left( \frac{n}{2^{d+1}} + O\left( n d^2 2^{-2d} \right)\right)$

via polymer/cluster-expansion methods (Collares et al., 28 Mar 2025).

For forbidden induced subgraphs $bK_a$ , sharp upper bounds on $I(G)$ are given by

$I(G) \leq n^{O(1)}\cdot \alpha^{O(\alpha)},$

with further improvements in $\chi$ -bounded families to $n^{O(1)}2^{O(\alpha)}$ (Bucić et al., 2024).

Probabilistic analysis addresses properties of the independent set sequence $(x_k(G))_{k=0}^{\alpha(G)}$ . For random trees and sparse random graphs, partial unimodality holds with high probability (Heilman, 2020).

5. Optimization and Algorithmics

The Maximum Independent Set (MIS) problem aims to compute $\alpha(G)$ . This problem is NP-hard in general, with various algorithmic frameworks developed for specific classes:

For cographs, distance-hereditary graphs, and cocomparability graphs, polynomial-time or linear-time dynamic programming or recurrence-based algorithms exist (Dyer et al., 2018, Chang et al., 2012).
In subclasses defined by forbidden induced subgraphs, e.g., $(P_4+P_4,\triangle)$ -free graphs, the structure theorem enables reduction to polynomially many complete bipartite subgraphs, yielding a poly-time algorithm (Mosca, 2020).
Exact enumeration of all independent sets of bounded size $k$ is FPT in $k$ for cocomparability and strongly orderable graphs (Dyer et al., 2018, Heinrich et al., 2021).

For large instances, heuristic evolutionary algorithms leveraging graph partitioning and local search (ARW, forced-insertion) scale to million-node graphs, outperforming classical local search (ARW) in many real-world families (Lamm et al., 2015).

6. Geometry and Spectral Theory

In algebraic and geometric extremal graph theory, the independence number of polarity graphs over projective planes, e.g., the Erdős–Rényi graph $ER_q$ , is governed by the Hoffman ratio bound: $\alpha(G) \leq \frac{-\lambda_{\min}}{d-\lambda_{\min}} n,$ with matching lower bounds $c q^{3/2}$ for $q$ a prime power (Tait et al., 2016). Extensions to orthogonal, unitary, and exotic polarities all support the conjecture $\alpha(G)=\Omega(q^{3/2})$ .

7. Independent-Set Graphs and Reconfiguration

For $k\geq 0$ , the $k$ -independent graph $I_k(G)$ is defined with vertices the independent sets of size at most $k$ , edges encoding adjacency by token addition/removal. For all $k$ , $I_k(G)$ is connected and bipartite, with maximum degree $n=|V(G)|$ . For $P_n$ , $|V(I_{\alpha}(P_n))|$ equals the $(n+2)$ nd Fibonacci number (Fatehi et al., 2015).

Reconfiguration problems, such as the Token Addition–Removal (TAR) model, are NP-complete in general but admit efficient algorithms in cographs and related classes.

References: