Paper: arXiv 2505.10373

Authors: Mattia Marzi, Francesca Giuffrida, Diego Garlaschelli, Tiziano Squartini

Abstract

The study of probabilistic models for the analysis of complex networks represents a flourishing research field. Among the former, Exponential Random Graphs (ERGs) have gained increasing attention over the years. So far, only linear ERGs have been extensively employed to gain insight into the structural organisation of real-world complex networks. None, however, is capable of accounting for the variance of the empirical degree distribution. To this aim, non-linear ERGs must be considered. After showing that the usual mean-field approximation forces the degree-corrected version of the two-star model to degenerate, we define a fitness-induced variant of it. Such a `softened’ model is capable of reproducing the sample variance, while retaining the explanatory power of its linear counterpart, within a purely canonical framework.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper is dense with advanced mathematics, including extensive use of probability theory, combinatorics, and statistical physics formalisms to derive non-linear constraints and model degeneracies. While it mentions real-world financial and social network applications and presents one illustrative figure, the focus is on theoretical derivations, approximation schemes, and model definitions rather than providing detailed empirical backtesting procedures, code, or datasets, making it more suited for theoretical exploration.

Research Flowchart

  flowchart TD
  A["Research Goal: Reproduce First & Second Moments of Empirical Degree Distributions"] --> B["Methodology: Define Fitness-Induced ERG Model"]
  A --> C["Data: Real-World Complex Network (Multi-Asset)"]
  B --> D["Computational Process: Mean-Field Approximation & Parameter Estimation"]
  C --> D
  D --> E["Outcome: Model Degenerates"]
  D --> F["Outcome: Softened Model Preserves Variance & Mean"]
  E --> G["Key Finding: Standard Mean-Field Fails for Non-Linear ERGs"]
  F --> H["Key Finding: Canonical Model Captures Degree Distribution Moments"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper defines a fitness-induced variant of the degree-corrected two-star model (fit2SM), a softened non‑linear ERG that can reproduce the sample variance while retaining explanatory power of its linear counterpart within a canonical framework.

Source sentence
After showing that the usual mean-field approximation forces the degree-corrected version of the two-star model to degenerate, we define a fitness-induced variant of it. Such a `softened' model is capable of reproducing the sample variance, while retaining the explanatory power of its linear counterpart, within a purely canonical framework.

Assumes

No explicit assumptions found in the text.

Data

  • Electronic Market for Interbank Deposits (eMID) transaction data market-prices unknown
    January 1999 to September 2012 · overnight interbank deposit segment
    Source sentence
    12 APPENDIX A. DATA DESCRIPTION AND TEMPORAL AGGREGATIONS We employ transaction-level records from the Electronic Market for Interbank Deposits (eMID), a screen-based market for unsecured deposits [8, 37, 38], restricting the analysis to the segment that represents the vast majority of the activity on such a market, i.e., the overnight one. Our sample spans trading days from January 1999 to September 2012: each

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • The fit2SM correctly reproduces the sample variance Var[k], whereas the UBCM overestimates it and the dcGM either over- or underestimates it. statistical
    Source sentence
    Graphical representation of the agreement between the sample variance Var[k] (black, dashed, vertical line) and its expected value ⟨Var[k]⟩under the UBCM (red), the dcGM (yellow), and the fit2SM (brown): while the UBCM steadily overestimates it and the dcGM either overestimates or underestimates it, the fit2SM correctly reproduces it.
  • The fit2SM accurately reproduces the empirical spectral radius, while the UBCM overestimates and the dcGM underestimates it. performance
    Source sentence
    In words, while the UBCM overestimates the spectral radius of the generative model (black, dashed, vertical line) and the dcGM underestimates it, the fit2SM correctly reproduces it.
  • Across daily, weekly, and monthly aggregations, the fit2SM yields smaller Bayesian Information Criterion (BIC) errors than the dcGM and the UBCM. performance
    Source sentence
    ots); • competes with the dcGM at the monthly time-scale (in fact, BICfit2SM < BICdcGM on 64.24% of monthly snapshots and BICfit2SM < BICUBCM on 33.94% of monthly snapshots but BICfit2SM < BICdcGM on 100.00% of monthly snapshots and BICfit2SM < BICUBCM on 97.78% of monthly snapshots since 2009); • compete with the dcGM at the quarterly time-scale during the last snapshots of our dataset (in fact, BICfit2SM < BICdcGM
  • In sparse configurations, the fit2SM outperforms the dcGM and competes with the UBCM by producing fewer isolated nodes in sampled configurations. performance
    Source sentence
    Figure 10, instead, compares the performances of the UBCM, the dcGM and the fit2SM in ‘producing’ configurations being characterized by a certain number of isolated nodes: as we have already observed, the sparser the configuration, the better the performance of the fit2SM - in these cases, the fit2SM outperforms the dcGM and compete with the UBCM on at least a portion of the dataset,

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 13 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 1 paper in the archive

Code

https://github.com/mattiamarzi/fit2SM — the authors' own repository

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (49)

  1. Unbiased sampling of network ensembles (2015) — link
  2. Enhanced gravity model of trade: Reconciling macroeconomic and network models (2019) — link
  3. Methodology for estimating the Dutch interfirm trade network (2019)
  4. Functional structure in production networks (2021) — link
  5. Reconstructing firm-level interactions in the Dutch input–output network (2022) — link
  6. Reconstruction methods for networks: The case of economic and financial systems (2018) — link
  7. Systemic risk analysis on reconstructed economic and financial networks (2015) — link
  8. Similarity-based link prediction in complex networks (2020) — link
and 41 more
  1. Monitoring supply networks from mobile phone data (2022) — link
  2. A machine learning approach to infer global supply chain disruption risks (2022) — link
  3. Local phase transitions in networks (2023) — link
  4. The role of the airline transportation network in the prediction and predictability of global epidemics (2006)
  5. Dynamical Processes on Complex Networks (2008)
  6. Networks: An Introduction (2010)
  7. Epidemic processes in complex networks (2015)
  8. Early-warning signals of topological collapse in interbank networks (2013) — link
  9. Complexity theory and financial regulation (2016) — link
  10. Pathways towards instability in financial networks (2017) — link
  11. Spectral signatures of structural change in financial networks — → 2409.03349
  12. Information theory and statistical mechanics (1957)
  13. Statistical mechanics of networks (2004)
  14. Maximum likelihood: Extracting unbiased information from complex networks (2008)
  15. Analytical maximum-likelihood method to detect patterns in real networks (2011)
  16. Randomizing bipartite networks: The case of the World Trade Web (2015) — link
  17. Maximum-entropy networks (2017)
  18. The statistical physics of real-world networks (2019) — link
  19. Statistical mechanics of the international trade network (2006)
  20. Entropy of network ensembles (2009)
  21. Statistical mechanics of the international trade network: Structural correlations and modeling (2012)
  22. A critical point for random graphs with a given degree sequence (1995) — link
  23. Generating uniformly distributed random networks (2006)
  24. Efficient and exact sampling of simple graphs with given arbitrary degree sequence (2010) — link
  25. A sequential importance sampling algorithm for generating random graphs with prescribed degrees (2011)
  26. Network-based prediction of unobserved edges in complex networks (2012)
  27. Unbiased degree-preserving randomization of directed binary networks (2012)
  28. The mathematical theory of minority games: Statistical mechanics of interacting agents (2009)
  29. Reconstructing Networks (2021)
  30. Solution of the two-star model of a network (2004)
  31. Fast and scalable likelihood maximization for exponential random graph models via GPU parallelization (2021) — link
  32. Epidemic spreading in scale-free networks (2001)
  33. DebtRank: Too central to fail? Financial networks, the FED and systemic risk (2012) — link
  34. Connected components in random graphs with given expected degree sequences (2002)
  35. Matrix Analysis (1985)
  36. Voter model on heterogeneous graphs (2005) — link
  37. Voter models on heterogeneous networks (2008) — link
  38. A network analysis of the Italian overnight money market (2008) — link
  39. Network Analysis of the e-MID Overnight Money Market: The Informational Value of Different Aggregation Levels for Intrinsic Dynamic Processes (2012)
  40. Social dynamics of financial networks (2018) — link
  41. Extracting the multi-timescale activity patterns of online financial markets (2018) — link

1 of 49 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.