Kleinberg nonprincipal hub-authority extraction
Algorithms AIMech:KleinbergNonprincipalHubAuthorityExtraction DRAFT
Aliases: Nonprincipal eigenvector hub-authority extraction
A link-analysis procedure that uses paired nonprincipal eigenvectors of the authority and hub Gram matrices of a query-focused directed page graph, selecting extreme coordinates to obtain additional hub and authority lists that implicitly summarize different linked topics.
Identity Notes & Concept Scope
Compared all 442 pinned AIO and 681 native identities before either cohort write; no normalized match. Ignored- and hidden-inclusive searches of three checkouts returned 419,78,76 lines. All 23 nonembedded unique payloads were read; six embedded matches were decoded and verified byte-identical to previously read search payloads. These were prior pending leads and contrasts, not existing maintained identities. Semantic assessment covered 127 AIO and 450 native definitions: 127/439 prior rows verified unchanged, eleven new native rows read. Fourteen complete AIO nodes/incident-edge/generated-record contexts and full PowerIterationClustering, LaplacianEigenmaps, KernelPCA, ChakrabartiFocusedCrawler and RelevanceWeightedTopicDistillation records were assessed. The original principal ranking and nonprincipal extraction were compared directly. The concept is the Section 6 nonprincipal paired-eigenvector extraction of additional hub/authority summaries, distinct from principal ranking, centered PCA and Laplacian embeddings. It does not supply a complete disjoint partition.
Claims & Grounded Evidence
5 Claims
Mechanism
#paired-modes
For directed adjacency A, consider nonprincipal eigenvectors x of A^T A and their corresponding hub directions in A A^T. For positive eigenvalue lambda and unit x, the unit paired vector is y=A x/sqrt(lambda). Applying the authority or hub propagation keeps these paired directions invariant up to scalar multiplication. The source uses modes beyond the principal pair to obtain additional strongly linked collections.
Mechanism
#coordinate-selection
For each selected pair, the source takes the c most positive and c most negative coordinates as hub and authority summaries. These are extrema-based representatives, not a full geometric embedding followed by a required partition. Sign reversal of both paired vectors exchanges the two ends without changing the represented pair of collections. When a mode has only one nonzero sign, the paper does not provide a meaningful second signed group or a zero-coordinate tie policy.
Limitation
#normalization-and-sign-boundaries
Proposition 6.1 prints y=A x although the earlier notation specifies unit eigenvectors. Positive modes require division by sqrt(lambda) for a unit pair; a zero mode maps to zero and cannot supply a nonzero paired eigenvector through this formula. The statement that every nonprincipal vector has both signs also needs qualification for reducible Gram matrices: an eleven-vertex example with nonzero eigenvalues 9 and 4 has a nonnegative secondary eigenvector. These are explicit curator algebraic qualifications, not claims about an inspected author implementation.
Scope
#implicit-cluster-scope
The extracted lists are intended as compact topic summaries of large clusters that are not explicitly represented. Different modes can repeat the same highly ranked pages, and the illustrated distinct collections were selected by hand from several modes. There is no guarantee of a disjoint partition, exhaustive query-sense coverage, unique semantic labeling or a prescribed number of distinct clusters. Bases within repeated eigenspaces are not uniquely determined by the eigenproblem.
Limitation
#historical-evaluation
The manuscript illustrates multiple senses and related communities using ranked URL lists and describes a narrow-query example selected by keyword counts among early nonprincipal-mode lists. These demonstrations do not establish universal semantic separation or measured clustering accuracy. The Section 8 human studies use enhanced systems and are not direct evidence for this extension alone; the sign of an eigenvector carries no inherent social or normative meaning.
⚠️ Curated Limitations & Negative Results
- The record denotes the Section 6 extension, not principal HITS ranking, all spectral graph partitioning, PCA or a trained model instance.
- The positive-mode normalization and reducible-matrix sign boundaries qualify broad printed statements. No author-code defect, repaired implementation or universal semantic guarantee is inferred.
- Mode selection, coordinate cutoffs, degeneracies and repeated leading pages affect summaries. Zero eigenvalues and one-sided modes do not provide the unqualified two-sided construction described in the source prose.
- Historical hand-selected examples and finite arithmetic do not certify recovery of real communities, exhaustive retrieval or current-web effectiveness. Enhanced-system user-study results remain separate.
- The full 34-page author manuscript and relevant rendered mathematics/results were inspected. No author software, historical graphs, live crawl or training was executed. Source pagination differs from journal metadata; DRAFT remains.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:Clustering
Clustering
|
related_match | The source explicitly treats additional hub/authority lists as implicit summaries of different linked topic clusters. This algorithm-to-task association retains a clustering relationship without asserting a complete partition, exact identity with the task or an AIO subclass axiom. |
Evidence Sources & Primary Literature
1 Source| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
kleinberg1999
|
Jon M. Kleinberg
|
primary paper
full_text
|
ID: doi:10.1145/324133.324140
sha256:a8503fe2c089…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-14T03:04:24Z
Updated: 2026-10-14T03:04:24Z