Clique-Stars in Graph Theory & Stellar Studies

• Clique-Stars are structures where clique-like and star-like patterns jointly determine properties in graph theory, privacy mining, Ramsey theory, algebra, and stellar clustering.
• They provide a framework for decomposing graphs via clique/double-star cutsets, leading to efficient algorithms and explicit bounds on optimization problems and chromatic numbers.
• In astrophysics, Clique-Stars capture hierarchical assemblies of young stars and the dispersal of stellar clusters, linking local star-centered structure with large-scale organization.

“Clique-Stars” is an Editor’s term for research settings in which clique-like and star-like structures jointly determine structure, algorithms, extremal behavior, or observed spatial organization. In graph theory, the term names a decomposition program built on clique cutsets and double-star cutsets for hereditary classes excluding most Truemper configurations (Boncompagni et al., 2017). In privacy-preserving graph mining, it denotes star-based local reports that support accurate $(p,q)$-clique enumeration under Local Differential Privacy (Sun et al., 2024). In Ramsey theory, it appears in problems contrasting cliques with stars, including both size Ramsey numbers for $K_{1,k}$ versus $K_n$ and 3-uniform hypergraph Ramsey numbers for $K_4^{(3)}$ versus $S_n^{(3)}$ (Miralaei et al., 2016, Conlon et al., 2022). In algebraic graph theory, it refers to the generalized star–clique operation on arithmetical structures and the induced transformation of critical groups (Diaz-Lopez et al., 2022). In stellar-population studies, the same label captures hierarchically clustered, clique-like assemblies of young stars across multiple spatial scales (Gouliermis et al., 2012, Davidge et al., 2011).

1. Clique cutsets, double-star cutsets, and hereditary graph classes

In the structural graph-theoretic setting, the central objects are holes, antiholes, and Truemper configurations: thetas, pyramids, prisms, and wheels. A wheel is a pair $(H,x)$, where $H$ is a hole and $x \in V(G)\setminus V(H)$ has at least three neighbors in $H$; a universal wheel has $x$ complete to $K_{1,k}$0, a twin wheel has $K_{1,k}$1 adjacent to exactly three consecutive vertices of $K_{1,k}$2 and to no other vertices of $K_{1,k}$3, and a proper wheel is any wheel that is neither universal nor twin. A clique cutset is a clique $K_{1,k}$4 such that $K_{1,k}$5 is disconnected. A star cutset is a cutset contained in $K_{1,k}$6 for one center $K_{1,k}$7, and a double-star cutset is a cutset $K_{1,k}$8 for adjacent centers $K_{1,k}$9 such that $K_n$0 is disconnected (Boncompagni et al., 2017).

The main ambient class is $K_n$1, the class of $K_n$2-free graphs, thus allowing possibly universal and twin wheels. Three subclasses are singled out: $K_n$3, which is $K_n$4-free; $K_n$5, which is $K_n$6-free; and $K_n$7, which is $K_n$8-free. Each class admits a clique-cutset decomposition into a basic class: $K_n$9 for $K_4^{(3)}$0, $K_4^{(3)}$1 for $K_4^{(3)}$2, $K_4^{(3)}$3 for $K_4^{(3)}$4, and $K_4^{(3)}$5 for $K_4^{(3)}$6. The corresponding decomposition theorems state that every graph in the class either belongs to its basic class or admits a clique cutset (Boncompagni et al., 2017).

The basic classes are described in terms of anticomponents, rings, hyperholes, hyperantiholes, and constrained low-$K_4^{(3)}$7 pieces. A hyperhole is obtained by blowing up a hole into a cyclic sequence of cliques, and a ring is a more general cyclic clique chain satisfying nested neighborhood conditions within each block. Rings are crucial atoms: every hyperhole is a ring, every hole in a ring picks exactly one vertex per block, removing any block $K_4^{(3)}$8 yields a chordal graph, and any wheel inside a ring is necessarily a twin wheel. This extends the chordal paradigm beyond hole-free graphs: chordal graphs admit Dirac’s decomposition “either complete or with a clique cutset,” whereas the present classes retain clique-based decompositions while allowing rings, hyperholes, hyperantiholes, and certain wheels (Boncompagni et al., 2017).

A decisive local restriction concerns vertex attachments to a hole. If $K_4^{(3)}$9 is a hole and $S_n^{(3)}$0 in a $S_n^{(3)}$1 graph, then either $S_n^{(3)}$2 is complete to $S_n^{(3)}$3, or $S_n^{(3)}$4 is a twin of some rim vertex, or $S_n^{(3)}$5 has at most two neighbors in $S_n^{(3)}$6 and they form a clique of two consecutive rim vertices. This restriction forces either ring structure or the existence of clique cutsets. Because no $S_n^{(3)}$7 and no wheel admits a clique cutset, the decomposition theorems also yield composition theorems: graphs in these classes can be obtained by repeatedly gluing basic atoms along cliques, except in the cap-free caveat noted for caps (Boncompagni et al., 2017).

2. Algorithmic consequences and $S_n^{(3)}$8-boundedness

The decomposition theory has direct algorithmic consequences. Recognition is polynomial and robust in all four classes: $S_n^{(3)}$9 is recognized in $(H,x)$0, $(H,x)$1 in $(H,x)$2, $(H,x)$3 in $(H,x)$4, and $(H,x)$5 in $(H,x)$6. The frameworks combine clique-cutset decomposition trees with atom recognition routines such as long-hole detection, ring detection in $(H,x)$7, and true-twin quotients for rings and hyperholes (Boncompagni et al., 2017).

Optimization problems split sharply by subclass. For maximum weight clique, the paper gives $(H,x)$8 algorithms for $(H,x)$9 and $H$0, and an $H$1 algorithm for $H$2, while maximum weight clique is NP-hard for $H$3. For maximum weight stable set, the stated bounds are $H$4 for $H$5, $H$6 for $H$7, and $H$8 for $H$9; the complexity for $x \in V(G)\setminus V(H)$0 remains open. For optimal vertex coloring, $x \in V(G)\setminus V(H)$1 admits an $x \in V(G)\setminus V(H)$2 algorithm and $x \in V(G)\setminus V(H)$3 admits an $x \in V(G)\setminus V(H)$4 algorithm, whereas coloring for $x \in V(G)\setminus V(H)$5 and $x \in V(G)\setminus V(H)$6 is open. The obstruction in $x \in V(G)\setminus V(H)$7 is explicit: coloring rings efficiently in polynomial time is unknown (Boncompagni et al., 2017).

The $x \in V(G)\setminus V(H)$8-boundedness theory is organized around small separators. The key double-star cutset theorem states that every $x \in V(G)\setminus V(H)$9 satisfies at least one of the following: $H$0 is cap-free, or $H$1 and $H$2 admits a double-star cutset $H$3 of size at most $H$4. This separator reduces $H$5 graphs to cap-free instances and feeds into polynomial $H$6-bounds via Ramsey theory and an isolation theorem for highly connected induced subgraphs (Boncompagni et al., 2017).

The resulting bounds are explicit: $H$7

$H$8

$H$9

A sharper asymptotic statement is also proved for $x$0: there exists $x$1 such that for $x$2,

$x$3

The $x$4 and $x$5 bounds are accompanied by explicit extremal constructions: the join of an odd hole and a clique gives $x$6 in $x$7, and the join of $x$8 copies of $x$9 gives $K_{1,k}$00 in $K_{1,k}$01 (Boncompagni et al., 2017).

3. Star-based local privacy for biclique counting

In privacy-preserving graph mining, Clique-Stars refers to the use of $K_{1,k}$02-stars as the local primitive for estimating biclique counts in a bipartite graph $K_{1,k}$03. A $K_{1,k}$04-clique is a complete bipartite subgraph $K_{1,k}$05, and its count is denoted $K_{1,k}$06. A $K_{1,k}$07-star centered at $K_{1,k}$08 is the event that $K_{1,k}$09 is connected to all nodes in some $K_{1,k}$10-subset $K_{1,k}$11 of its neighbors across the bipartite boundary. The key combinatorial identity is that $K_{1,k}$12 counts can be written as sums of binomial terms over shared neighborhoods, so observing $K_{1,k}$13-stars allows estimation of the common-neighborhood sizes needed to recover $K_{1,k}$14 (Sun et al., 2024).

The privacy model is $K_{1,k}$15-LDP. Each user constructs a $K_{1,k}$16-stars neighboring list $K_{1,k}$17 locally and perturbs each binary entry with Warner’s randomized response, where $K_{1,k}$18. The mechanism is two-round. In round 1, users send RR-noised $K_{1,k}$19-stars lists to the collector. In round 2, users compute motif counts using only noisy data returned by the collector and send back final counts; because this second stage is post-processing, it does not consume additional privacy budget (Sun et al., 2024).

For the $K_{1,k}$20-stars LDP estimator, the paper uses noisy local counts $K_{1,k}$21 of $K_{1,k}$22-cliques and $K_{1,k}$23 of $K_{1,k}$24-cliques, with debiasing

$K_{1,k}$25

An absolute value correction replaces $K_{1,k}$26 by

$K_{1,k}$27

motivated by the fact that unbiased correction can be negative in sparse graphs. A further $K_{1,k}$28-stars sampling technique retains each RR-noised star with probability $K_{1,k}$29 to reduce cost and prune spurious stars (Sun et al., 2024).

The theoretical guarantees are exact. Both the edge-LDP baseline and the $K_{1,k}$30-stars estimator are unbiased: $K_{1,k}$31 If $K_{1,k}$32, then

$K_{1,k}$33

whereas

$K_{1,k}$34

The leading exponent drops from $K_{1,k}$35 to $K_{1,k}$36, which is the main utility advantage. The privacy comparison theorem states that if a local mechanism provides $K_{1,k}$37–$K_{1,k}$38-stars LDP, then it also provides $K_{1,k}$39 edge LDP (Sun et al., 2024).

Empirically, the method is evaluated on Gplus, IMDB, GitHub, and Facebook, using relative error and $K_{1,k}$40 loss. Across all datasets and settings, $K_{1,k}$41-stars LDP outperforms edge LDP in relative error and $K_{1,k}$42 loss. The paper reports that absolute value correction reduces $K_{1,k}$43 loss by up to $K_{1,k}$44 orders of magnitude in sparse settings, and that $K_{1,k}$45-stars sampling with $K_{1,k}$46 cuts relative error by about three orders of magnitude (Sun et al., 2024).

4. Ramsey-theoretic clique–star phenomena

Clique–star interactions also appear in Ramsey theory. For graphs, the size Ramsey number $K_{1,k}$47 is the minimum number of edges of a graph $K_{1,k}$48 such that every red-blue coloring of $K_{1,k}$49 yields a red copy of $K_{1,k}$50 or a blue copy of $K_{1,k}$51. The star is $K_{1,k}$52, and the clique is $K_{1,k}$53. Faudree and Sheehan conjectured that

$K_{1,k}$54

but Pikhurko showed that this conjecture is false for $K_{1,k}$55 and $K_{1,k}$56. Miralaei, Omidi, and Shahsiah proved that for a given $K_{1,k}$57 and $K_{1,k}$58, the equality $K_{1,k}$59 does hold, thereby confirming the Faudree–Sheehan formula in this large-$K_{1,k}$60 regime (Miralaei et al., 2016).

In 3-uniform hypergraphs, the star $K_{1,k}$61 on $K_{1,k}$62 vertices has all $K_{1,k}$63 triples incident to a fixed center. The off-diagonal Ramsey number $K_{1,k}$64 satisfies

$K_{1,k}$65

for positive constants $K_{1,k}$66. The paper emphasizes that this is an unusual intermediate growth regime: both bounds are super-polynomial in $K_{1,k}$67 but sub-exponential in $K_{1,k}$68, unlike the more familiar behavior of many off-diagonal hypergraph Ramsey numbers against cliques (Conlon et al., 2022).

The upper bound is obtained through a grid-graph Ramsey problem on $K_{1,k}$69: determine the least $K_{1,k}$70 such that every 2-edge-coloring contains either a red rectangle or a blue $K_{1,k}$71. An explicit correspondence maps horizontal and vertical grid edges to triples of a bipartite 3-graph $K_{1,k}$72, under which a red rectangle corresponds to a red $K_{1,k}$73 and a blue row- or column-clique corresponds to a blue $K_{1,k}$74. The lower bound uses a layered pseudorandom construction on a grid with no rectangles and controlled row/column sparsity (Conlon et al., 2022).

For larger clique size, the phenomenon changes: for each fixed $K_{1,k}$75, there exist constants $K_{1,k}$76 such that

$K_{1,k}$77

This isolates $K_{1,k}$78 as the pivotal case producing the intermediate regime (Conlon et al., 2022).

5. Generalized star–clique operations and critical groups

In algebraic graph theory, clique–star interaction is encoded by the Keyes–Reiter generalized star–clique operation on arithmetical structures. An arithmetical structure on a finite, connected graph without loops is a pair $K_{1,k}$79 satisfying

$K_{1,k}$80

with $K_{1,k}$81 primitive under the normalization $K_{1,k}$82. The associated matrix $K_{1,k}$83 has rank $K_{1,k}$84, and the cokernel decomposes as

$K_{1,k}$85

where $K_{1,k}$86 is the critical group (Diaz-Lopez et al., 2022).

The generalized star–clique operation is performed at a chosen vertex $K_{1,k}$87. The new graph $K_{1,k}$88 deletes $K_{1,k}$89, scales every edge not incident to $K_{1,k}$90 by $K_{1,k}$91, and adds $K_{1,k}$92 parallel edges between each pair of neighbors $K_{1,k}$93 of $K_{1,k}$94. The updated parameters are

$K_{1,k}$95

At the matrix level,

$K_{1,k}$96

so $K_{1,k}$97 is obtained from $K_{1,k}$98 by pivotal condensation along the last row and last column (Diaz-Lopez et al., 2022).

This matrix identity drives the critical-group analysis. If $K_{1,k}$99 is the gcd of all $K_n$00 minors of $K_n$01, then for all $K_n$02,

$K_n$03

and consequently

$K_n$04

where $K_n$05 is the gcd of the last row of $K_n$06. For critical-group orders this yields

$K_n$07

Both bounds are sharp in the non-simple case (Diaz-Lopez et al., 2022).

The invariant factors are exact for simple graphs, equivalently when $K_n$08. If

$K_n$09

then

$K_n$10

Hence

$K_n$11

and $K_n$12. In the classical case $K_n$13, the critical group is preserved (Diaz-Lopez et al., 2022).

6. Hierarchical stellar “cliques” and dispersing star complexes

In stellar-population studies, Clique-Stars denotes clustered young stars arranged in nested, clique-like groupings across scales. In the NGC 346/N66 region of the Small Magellanic Cloud, HST/ACS observations over a field of about $K_n$14 pc were analyzed with stellar density maps, Cartwright and Whitworth’s $K_n$15 parameter, dendrograms, and the $K_n$16-variance wavelet transform. The stellar density maps were built on a regular grid with pixel size $K_n$17 and smoothed with a Gaussian filter of FWHM $K_n$18 pixels $K_n$19. The central dominant aggregate is outlined by the $K_n$20 isopleth, and NGC 346 proper emerges around $K_n$21 (Gouliermis et al., 2012).

The $K_n$22 parameter combines the mean edge length of the minimum spanning tree and the mean pairwise separation, with $K_n$23 indicating hierarchical or fractal substructure. For NGC 346/N66, the MST analysis yields a $K_n$24 that maps to a fractal dimension $K_n$25. The dendrogram reveals multiple levels of hierarchy concentrated inside the central aggregate outlined by the $K_n$26 isopleth. The $K_n$27-variance shows a power-law regime over lags $K_n$28 to $K_n$29, a best-fit spectral index $K_n$30, $K_n$31, $K_n$32, and a break in self-similarity at $K_n$33 pixels, i.e. $K_n$34 $K_n$35 pc). Together these diagnostics indicate strong hierarchical, fractal-like clustering (Gouliermis et al., 2012).

A related but dynamically different picture appears in the outer disk of M33. There the star–star separation function (S3F), defined as the histogram of all pairwise angular separations in an age-selected sample, was applied to bright main-sequence stars near the northern edge of the disk at $K_n$36 kpc. The 10 Myr sample shows a strong excess for separations $K_n$37 $K_n$38 kpc), with the width of the low-$K_n$39 peak indicating grouping scales $K_n$40 $K_n$41 pc), and a second peak at $K_n$42 $K_n$43 kpc) caused by beating between two large young stellar complexes in the field (Davidge et al., 2011).

By 40 Myr, clustering signatures broaden and weaken, with most stars separated by up to at least $K_n$44 $K_n$45 kpc). By 100 Myr, the S3F is nearly flat between $K_n$46 and $K_n$47 $K_n$48–$K_n$49 kpc), with no excess at $K_n$50 and a minimum typical separation of $K_n$51 $K_n$52 kpc). The observed redistribution is consistent with random velocities of order $K_n$53, which imply $K_n$54 kpc over $K_n$55 yr. The paper further connects the disappearance of clustering by $K_n$56 Myr to cluster disruption, estimating characteristic birth masses of $K_n$57–$K_n$58 from the disruption law $K_n$59 with $K_n$60 and an inner-disk baseline $K_n$61 Gyr for a $K_n$62 cluster (Davidge et al., 2011).

Taken together, these astronomical uses of Clique-Stars describe two related phenomena: hierarchical clique-like assembly of young stars across scales in NGC 346/N66, and the rapid dispersal of initially compact, sub-kpc stellar groupings in the outer disk of M33. A plausible implication is that the term consistently marks settings in which local star-centered structure and larger clique-like organization are not alternatives but coupled descriptions of the same underlying system (Gouliermis et al., 2012, Davidge et al., 2011).