REVIEW 3 major objections
Iterated renormalisation of inhomogeneous random graphs converges to a two-parameter attractor family, with different limits in light- and heavy-tailed regimes.
Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →
T0 review · grok-4.5
2026-07-15 05:55 UTC pith:JX2EPUAK
load-bearing objection Abstract-only claim of a dual-regime renormalisation attractor for inhomogeneous graphs; coherent programme, but no proofs to inspect. the 3 major comments →
Renormalisation of Inhomogeneous Random Graphs
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
When an inhomogeneous random graph with i.i.d. vertex weights and suitably scaled connection functions is subjected to iterated greedy block renormalisation, the sequence of renormalised graphs converges to a two-parameter family that is an attractor: a homogeneous Erdős–Rényi graph in the light-tailed regime and an inhomogeneous graph with stable infinite-mean weights and exponential disconnection function in the heavy-tailed regime.
What carries the argument
The renormalisation map that aggregates vertices into equal-sized blocks via a greedy algorithm and defines an aggregated edge (respectively self-loop) whenever at least one original edge crosses the cut (respectively lies inside the block); iterated application of this map, under regime-specific scalings of the connection functions, produces the claimed convergence.
Load-bearing premise
The connection functions and the law of the vertex weights must be scaled in a regime-specific way so that the renormalisation map stays inside the claimed basin of attraction; without those scalings the iteration need not converge to the stated two-parameter family.
What would settle it
Choose a concrete connection function and weight law that the paper places in one regime, apply the stated scaling and the greedy aggregation repeatedly on large finite graphs, and check whether the empirical edge probabilities and weight statistics converge to the predicted Erdős–Rényi or stable-weight exponential form; systematic deviation after many iterations would falsify the attractor claim.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript studies inhomogeneous random graphs with i.i.d. vertex weights, edges present independently with probability given by a bivariate function of the endpoint weights, and self-loops present independently with probability given by a univariate function of the vertex weight. A renormalisation map is defined by greedy aggregation of vertices into equal-sized blocks: an aggregated edge exists between distinct blocks iff at least one cross-edge is present, and an aggregated self-loop exists iff at least one internal self-loop or internal edge is present. The central claim is that, starting from appropriately scaled connection functions, iterated application of this map converges to a two-parameter family of random graphs that acts as an attractor in a universality class. Two regimes are distinguished: a light-tailed regime whose limit is a homogeneous Erdős–Rényi graph, and a heavy-tailed regime whose limit is an inhomogeneous graph with stable infinite-mean weights and an exponential disconnection function. Different scalings are required in each regime; which regime prevails depends on the connection functions and the weight law.
Significance. If the claimed convergence and attractor properties are established rigorously, the work would supply a renormalisation-group description of inhomogeneous random graphs under greedy coarse-graining, with an explicit two-parameter universal family and a clean light-tailed/heavy-tailed dichotomy. The identification of a homogeneous ER attractor in the light-tailed regime and of a stable infinite-mean weight law with exponential disconnection in the heavy-tailed regime would be a substantial contribution to the probabilistic theory of complex networks and to universality for discrete structures. The programme is mathematically precise at the level of the abstract and cleanly separates the two regimes.
major comments (3)
- Only the abstract is available for review. The load-bearing claim that iterated renormalised graphs converge to a two-parameter attractor family (light-tailed ER vs heavy-tailed stable infinite-mean weights with exponential disconnection) cannot be assessed without the proofs, the topology of convergence, error estimates, and technical lemmas. This is an information gap, not an identified internal inconsistency.
- Abstract: the iteration is asserted to remain inside a basin of attraction only after regime-specific scalings of the connection functions and of the weight law (“starting from appropriately scaled connection functions”; “Different scalings are needed for the two regimes”). Without the full text it is impossible to verify that those scalings are well-defined, that the greedy aggregation map is continuous in the relevant topology, or that the stated two-parameter family is attractive under iteration.
- Abstract: the criterion that selects the light-tailed versus heavy-tailed regime is said to depend on the connection functions and the law of the weights, but the precise separation condition and the form of the required scalings are not available for inspection. Any correctness assessment of the dual-regime claim therefore remains provisional.
Circularity Check
No significant circularity: abstract-only pure mathematical derivation of renormalisation attractors under stated model assumptions.
full rationale
Only the abstract is available. It states a pure mathematical claim: starting from i.i.d. vertex weights, independent edges given weights, and a greedy aggregation renormalisation map, iterated renormalised graphs converge (under regime-specific scalings of the connection functions) to a two-parameter attractor family—homogeneous Erdős–Rényi in the light-tailed regime and an inhomogeneous graph with stable infinite-mean weights plus exponential disconnection in the heavy-tailed regime. No data fitting, no parameter estimation from observations, no self-definitional identities, and no load-bearing uniqueness theorems imported from the authors’ prior work appear in the abstract. The phrase “appropriately scaled connection functions” and the remark that “Different scalings are needed for the two regimes” are ordinary modelling hypotheses that keep the map inside a claimed basin of attraction; they do not make the claimed limit equal to its input by construction. Because the full text is unavailable, no equation-level reduction can be exhibited, and the honest finding under the given material is therefore score 0 with an empty steps list. Any residual concern is an information gap about continuity of the aggregation map or attractiveness of the fixed family, not circularity.
Axiom & Free-Parameter Ledger
axioms (4)
- domain assumption Vertex weights are i.i.d.; edges (and self-loops) are independent given the weights, with probabilities given by fixed uni- and bi-variate connection functions.
- domain assumption The renormalisation map aggregates vertices into equal-size groups via a greedy algorithm and defines aggregated edges/self-loops by the existence of at least one original edge/self-loop.
- ad hoc to paper Connection functions and weight laws are scaled in a regime-dependent manner so that the iteration remains inside the claimed basin of attraction.
- standard math Standard measure-theoretic probability (existence of i.i.d. sequences, independence, convergence in distribution on graph spaces).
read the original abstract
We consider inhomogeneous random graphs in which vertices are assigned i.i.d.\ random weights, pairs of distinct vertices are connected by an edge independently with a probability that is a bi-variate function of the weights of the vertices, and single vertices are connected to themselves by a self-loop independently with a probability that is a uni-variate function of the weight of the vertex. We apply a renormalisation transformation in which vertices are aggregated into groups of equal size according to a greedy algorithm, namely, distinct groups of aggregated vertices are connected by an aggregated edge if and only if there is at least one edge connecting two constituent vertices across the groups, while a group of aggregated vertices is connected to itself by an aggregated self-loop if and only if there is at least one self-loop at an internal vertex or one edge connecting a pair of distinct internal vertices. We analyse what happens when the renormalisation transformation is iterated. In particular, we show that, starting from appropriately scaled connection functions, the iterated renormalised graphs converge to a two-parameter family of random graphs, acting as an attractor in a universality class. We consider a light-tailed regime, for which the scaling limit is a homogeneous Erd\H{o}s--R\'enyi random graph, and a heavy-tailed regime, for which the scaling limit is an inhomogeneous random graph with stable infinite-mean random weights and an exponential disconnection function. Different scalings are needed for the two regimes. Which of the two regimes prevails depends on the choice of the connection functions and the choice of the law of the random weights.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.