Zhou random-walk distance-profile dissimilarity
Mathematical Methods AIMech:ZhouRandomWalkDissimilarity DRAFT
Aliases: Zhou neighbor dissimilarity index
A pairwise graph dissimilarity that compares the mean first-passage-time profiles of two adjacent vertices by taking the Euclidean norm of their differences to all other vertices and dividing that norm by N-2, for a graph of N vertices.
Identity Notes & Concept Scope
Compared the pairwise functional against all 442 pinned AIO and 698 native identities; no normalized match. Ignored- and hidden-inclusive searches of three full checkouts returned 62,0,0 lines, read as the previous Zhou record/report, source evidence, query strings or pending leads. Semantic assessment covered 155 AIO and 543 native definitions: 155 AIO and542 prior native rows verified unchanged, plus the new attractor-closure row and full record read. Thirteen complete AIO contexts and six full native neighbors assessed. The identity is the symmetric nonnegative pairwise profile score, not first-passage distance itself, a node centrality, original attractor closure or the full subsequent hierarchy. Distinct vertices can have zero score and the original domain is adjacent pairs.
Claims & Grounded Evidence
6 Claims
Scope
#walk-and-profile-domain
The maintained domain is a finite connected undirected graph with N>2, symmetric nonnegative adjacency weights and positive vertex strengths. A walker moves from i to j with probability A_ij/sum_l A_il. For each distinct adjacent pair (i,j), the functional compares d_ik and d_jk for every k other than i and j, where d is the mean first-hitting-time matrix on the entire input graph. Unweighted graphs are a special case. The endpoint coordinates are omitted, so diagonal self-return conventions do not affect this index.
Mechanism
#exact-source-normalization
In the inspected arXiv v1, Lambda(i,j)=sqrt(sum_{k not in {i,j}}(d_ik-d_jk)^2)/(N-2). The denominator lies outside the square root. Root-mean-square normalization would divide the same norm by sqrt(N-2), yielding a value sqrt(N-2) times larger. On a four-cycle the source index squared is 1/2, whereas the root-mean-square value squared is 1. No equivalence between these numerical threshold scales is asserted.
Limitation
#symmetric-nonseparating-score
The index is symmetric and nonnegative even though its input passage times are generally asymmetric. It need not be positive for different vertices: two vertices with identical hitting profiles to all remaining vertices have value zero, including any pair in an unweighted complete four-vertex graph. The source defines it on adjacent pairs; extending it to all pairs or asserting metric separation or a triangle inequality requires additional justification.
Mechanism
#global-dynamics-not-local-overlap
The compared profiles depend on whole-graph random-walk passage, not only shared-neighbor counts, shortest-path lengths or stationary visit probabilities. Multiplying every adjacency weight by one common positive factor leaves the transition matrix, all mean passage times and the index unchanged. Computing full-network distances before later subdivision is distinct from recomputing them within each induced subgraph.
Limitation
#hierarchy-algorithm-is-separate
The source’s separate hierarchy procedure thresholds these indices into friend sets and simultaneously reassigns unstable vertices by total interaction. This functional alone defines neither a partition nor guaranteed hierarchy termination. A curator complete four-vertex weighted graph with within-pair weights 1 and cross-pair weights 2 has zero within-pair indices and cross-pair index squared 25/288. A threshold below the latter index yields two friend pairs; each node has own-group weight 1 and other-group weight 4. Simultaneous moves swap all group labels, leaving every node unstable and preventing the printed step-4 stopping condition. This is a scoped counterexample to unconditional termination of that printed adjustment rule, not an execution of author software.
Limitation
#interpretation-and-version-limits
The source interprets small index values as similar graph perspectives and uses them in historical synthetic, social and yeast-network clustering examples. Its performance and protein-function hypotheses concern the full hierarchy pipeline and specific data, not independent validation of this pairwise functional or causal biology. This record follows the inspected arXiv v1 normalization; journal-byte equivalence and version-specific numerical changes were not established.
⚠️ Curated Limitations & Negative Results
- The maintained identity is the adjacent-pair dissimilarity functional, not the complete threshold hierarchy, attractor communities, a learned graph representation or a node centrality.
- N>2 and connected positive-strength input are explicit. Disconnected cross-component passage, zero denominators and directed generalizations are outside scope.
- The source denominator is outside the radical. Numerical thresholds cannot be copied from a root-mean-square convention without rescaling.
- Tie policy, friendless vertices, exhausted thresholds and synchronous reassignment termination require separate algorithm assessment. The finite cycling example preserves the source’s simultaneous update rule.
- All ten source pages were read as text, with the defining equation and algorithm steps inspected visually; other displays remain text-only. No author code, raw benchmark replication, journal-byte equivalence or biological hypothesis validation. DRAFT remains.
Artificial Intelligence Ontology (AIO) Grounding
Mapped AIO Classes
| AIO Target Class | Mapping Relation | Mapping Rationale |
|---|---|---|
AIO:MathematicalFunction
Mathematical Function
|
broad_match | The local concept assigns a nonnegative number to a graph and an adjacent vertex pair using an explicit equation. AIO MathematicalFunction is the broader rule-to-value class; the local identity is a functional rather than a particular solver. |
AIO:MarkovChain
Markov Chain
|
related_match | The compared profiles are mean passage times of a specified Markov walk. This relates the pairwise functional to its generating process, without asserting process equivalence or inheriting the target’s Network/hidden-layer context. |
Evidence Sources & Primary Literature
2 Sources| Source ID | Publication / Document | Type / Access | Reference & Checksum |
|---|---|---|---|
zhou2003dissimilarity
|
Haijun Zhou
|
primary paper
full_text
|
ID: arxiv:physics/0302032v1
sha256:820fabdced2d…
|
zhou2003attractors
|
Haijun Zhou
|
primary paper
full_text
|
ID: arxiv:physics/0302030v1
sha256:2da11ce6cf7e…
|
Curation Provenance
Curator: Codex
Method: ai_assisted
Model: GPT-6
Created: 2026-10-14T08:02:02Z
Updated: 2026-10-14T08:02:02Z
Source File:
data/curated/ZhouRandomWalkDissimilarity.yaml