Pith. sign in

REVIEW 5 major objections 3 minor 41 references

Scale-free Points-of-Interest Distribution in a City Emerging from Homogeneous Poissonian-point Processes

T0 review · 5 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper shows that Bologna's power-law POI distribution is a mathematical consequence of aggregating locally homogeneous Poisson processes, with the exponent set by a simple area-intensity scaling identity.

desk verdict The paper's central exponent relation is wrong: the correct tail is α=β, not β−1/2, so the claimed generative mechanism and the validation that depends on it collapse, though the question deserves a careful referee. read the letter →

arxiv 2509.01699 v1 pith:GVQ3YVNY submitted 2025-09-01 physics.soc-ph

Eleonora Andreotti , Ulysse Marquis , Maurizio napolitano , Riccardo Gallotti This is my paper · ORCID
classification physics.soc-ph MSC 60G5562E20
keywords power-lawdistributionsPoissonpointprocessespointsofinteresturbanscalingspatialheterogeneitymixturemodelshierarchicalDBSCANBologna
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the heavy-tailed, power-law-like distribution of commercial points of interest (POIs) across a city can be produced entirely by layering locally ordinary random processes, with no need for self-organized criticality or other global organizing mechanisms. Its central derivation shows that if the urban surface is decomposed into regions whose average POI count per fixed-size cell is an integer j and whose total areas shrink like j^{-β}, then the city-wide count distribution decays as a power law with exponent β - 1/2. Because the regions are defined by shared intensity rather than spatial contiguity, the mechanism tolerates fragmented urban geography. The authors verify the required conditions in Bologna data by re-clustering unit cells into power-law-sized regions, and they extend the framework to Poisson mixtures so the strict area-intensity coupling can be relaxed. If correct, the work converts a widely observed urban scaling pattern from a mystery into a derivable consequence of local regularity plus heterogeneity.

What carries the argument

The engine is the area-weighted Poisson mixture over intensity-indexed regions: P(X=k) = (1/A_S) Σ_j A_j · Pois(k; λ_j). The paper imposes the coupling λ_j = j and A_j ~ j^{-β}; the resulting series is evaluated by a saddle-point argument using Robbins' factorial bounds, in which only terms near j ≈ k survive, yielding the closed-form decay exponent α = β - 1/2. The same series identity reappears in the Poisson mixture model and in the hybrid hierarchical model, where the fixed λ_j is replaced by an intra-region mixture of intensities; the hybrid model derives its tail by applying the same asymptotic estimate to area-weighted mixture masses near intensity k.

What would settle it

Compute, for a given city and POI category, the empirical regions obtained by sorting unit cells by count and grouping them into equal-sized bins whose areas follow a chosen power law, then fit log(area) against log(bin index) and regress the estimated Poisson intensity on bin index. If the area slope is not -β ≈ -(α+1/2) or the intensity is not linear with scatter comparable to the Bologna fits, the derivation's assumptions fail. A cleaner test: simulate the same Poisson mixture with A_j and λ_j drawn independently; if a power-law tail with the predicted exponent still appears, the mechanism

Watch

Extended reading notes

Core claim

The central claim is a scaling identity. Let X be the number of POIs in a fixed-size ball. If the city is partitioned into regions indexed by j, where each region is internally homogeneous with Poisson intensity λ_j = j and occupies total area A_j ~ j^{-β}, then P(X=k), the area-weighted Poisson mixture, is asymptotically a power law: P(X=k) ~ k^{-(β - 1/2)}. The argument uses Robbins' bounds on factorials to show the series Σ_j j^{k-β} e^{-j}/k! is dominated by terms with j ≈ k; the peak window contains Θ(k) terms, each of size O(k^{-β-1/2}), leaving O(k^{-β+1/2}). The paper also shows that this power law inherits only the joint distribution of areas and intensities, not their spatial arran

Load-bearing premise

The load-bearing assumption is the imposed coupling that regions can be ordered so that their Poisson intensities equal successive integers j while their total areas shrink as j^{-β}; this scaling is assumed for mathematical tractability rather than derived from any urban process, and the paper acknowledges that natural DBSCAN clusters do not show it.

Editorial extensions

If this is right

  • Observing a power law in POI counts can no longer be read as evidence of criticality or preferential attachment; the same shape follows from uniform local randomness with heterogeneous intensity levels.
  • The measured exponent α directly estimates the area-intensity coupling β = α + 1/2, giving a quantitative handle on urban structure from a single fitted number.
  • Because contiguity is irrelevant, polycentric and fragmented cities can still show clean scaling, so the model applies beyond monocentric urban forms.
  • The hybrid Poisson-mixture formulation lets the generative mechanism absorb local deviations, making it usable as a null model or classifier for urban regions.
  • Re-clustering by ascending count into power-law-sized bins provides a direct empirical check: real categories show linear intensity growth (R² ≈ 0.99), so the derivation is not purely formal.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same area-intensity superposition should, by symmetry, generate heavy tails for any count variable defined over fixed windows of a heterogeneous Poisson field—for example, taxi pickups or service requests—provided the region-level coupling A_j ∝ j^{-β} holds; testing this on mobility data would separate the general mechanism from POI-specific economics.
  • The finite-size convergence behavior reported here implies that empirical exponents estimated on small or weakly contrasted cities will be biased upward for low β; cross-city comparisons of α should therefore be corrected for the effective number of intensity regions.
  • One could turn the mechanism into a temporal diagnostic: if mixture weights shift over time, changes in the fitted tail exponent trace commercial densification or decline before aggregate counts change.
  • A direct falsifying experiment is to re-fit the same model on a second city and check whether the recovered relation between region area and intensity has the same power-law form; if not, the Bologna validation may reflect a favorable one-off rather than a universal mechanism.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 3 minor

Summary. The paper proposes that heavy-tailed distributions of urban Points of Interest (POIs) can emerge from the aggregation of locally homogeneous Poisson point processes. The theoretical model assumes a surface partitioned into regions with areas A_j ~ j^{-β} and Poisson intensities λ_j = j, and derives a predicted global power-law exponent α = β − 1/2. The authors test this on a synthetic surface and on Foursquare data for Bologna, and introduce a hybrid hierarchical model combining spatial clusters with Poisson mixtures. They report local Poisson behavior in DBSCAN clusters and claim that re-clustering the city into regions with power-law areas recovers the model's assumptions, providing empirical validation.

Significance. If the derivation were correct, the paper would give a simple, non-criticality generative mechanism for urban scaling, with a concrete exponent relation and a falsifiable synthetic test. The paper is candid about the strength of its assumptions, and the hybrid mixture extension is a reasonable modeling idea. However, the central asymptotic calculation is wrong: the correct exponent is α = β, not β − 1/2. The synthetic experiment actually supports the corrected relation, and the empirical validation in Sec. IV B is circular by construction. The quantitative claims of the paper are therefore not supported in their current form. The qualitative insight—that aggregating locally regular Poisson processes can produce heavy tails—remains plausible, but the paper's specific theoretical and empirical contributions need substantial correction.

major comments (5)
  1. [Sec. III A 1, Eq. (3)] The summation overcounts the contributing terms. The Poisson kernel e^{-j} j^k/k! is concentrated in a window of width O(√k) around j = k, not in a Θ(k) interval. For j = k + y, the summand is approximately k^{-β} e^{-y^2/(2k)} / √(2πk); summing over y = O(√k) gives S_k = Θ(k^{-β}), hence α = β. The claimed α = β − 1/2 is therefore incorrect, and this error propagates to Secs. III C, IV A, and IV B.
  2. [Sec. IV A, Figs. 7–8] The reported synthetic fit for β = 2.5 gives α ≈ 2.7, which is close to β = 2.5 and far from the claimed β − 1/2 = 2.0. The text attributes this 0.7 discrepancy to finite size, but the discrepancy is in the wrong direction and is instead consistent with the corrected exponent α = β. The 'expected' curves in Fig. 8 should be recomputed with the corrected relation.
  3. [Sec. IV B] The empirical validation is circular. The authors estimate α from the data, define region sizes as j^{-(α+1/2)}, sort balls by POI count, and assign them sequentially to regions. This guarantees that average counts increase with region index; it is not an independent test of the model. Moreover, under the corrected α = β relation, the area exponent should be β = α, not α + 1/2. The claim of 'striking empirical validation' is not justified.
  4. [Sec. III C] The hybrid-model tail estimate is internally inconsistent with Sec. III A. If the aggregate weight of components with λ ~ k satisfies W_k ~ k^{-β}, then the total contribution is W_k · O(k^{-1/2}) = O(k^{-β-1/2}), i.e., α = β + 1/2, not the α = β − 1/2 derived earlier. The section invokes the erroneous 'previous asymptotic estimate' and needs to be reconciled after the central derivation is corrected.
  5. [Sec. IV A, Fig. 7] The reported KS p-value is 0.0625, below the 0.1 threshold the paper itself adopts in Sec. II B. By the authors' own criterion, the power-law fit to the synthetic data is rejected. The statement that the distribution 'does not significantly deviate' from a power law is therefore not supported by the reported statistic.
minor comments (3)
  1. [Sec. III A 1] The symbol β is used both for the area exponent and for the ratio j/k in the critical-regime analysis. Rename one of them (e.g., use r for j/k) to avoid confusion.
  2. [Fig. 7 and Fig. 8 captions] The caption of Fig. 7 reads 'Theoretical Power Law, =2.70' and should read 'α = 2.70'. Also, '10 6 clusters' and '≈ 300' should be typeset as 10^6 and ≈300.
  3. [Sec. III C] There is a typo: 'Subbsec-tion' should be 'Subsection'. Also, the cross-references to 'Section III C' from within Section III C are confusing and should be fixed.

Circularity Check

2 steps flagged · score 6.0 of 10

Sec. IV B validation is circular: the fitted exponent α is used to construct the region areas (β=α+1/2), so 'recovering' the area power law is a tautology and the linear intensity growth is an order-statistic artifact of the same fitted distribution.

  1. self definitional [Section IV B (Empirical Validation of the Poisson Mechanism), paragraph beginning 'Here, we demonstrate...']
    "Specifically, we first compute the scaling parameter α of the global distribution for the category and predefine the number of regions. Each region is then composed of a number of unitary balls proportional to the region index raised to the power of −(α + 1/2)."

    The area scaling exponent is set to β = α + 1/2 using the data-fitted α and the paper's own relation α = β − 1/2. Consequently, the region areas are constructed to follow j^{-(α+1/2)}. When the paper later states that it 'recovers all the assumptions required for the generation of a power-law distribution with exponent α, namely ... region areas that follow a power law with exponent β', this is not an empirical finding but a restatement of the construction. The model's input is the fitted α, and the 'recovery' simply returns it.

  2. fitted input called prediction [Section IV B, paragraph following the construction and Fig. 9]
    "We sort the unitary balls in ascending order according to the number of POIs contained in each, and assign them sequentially to the regions. Fig. 9 shows ... how the Poissonian intensity measured in each region grows linearly with the region index, as it is expected from our model. This provides striking empirical validation of the theoretical result."

    Because the balls are sorted by the response variable (POI count) and the region sizes are fixed from the previously fitted α, the group-wise mean intensities are order statistics of the very empirical distribution used to estimate α. The near-linear growth in Fig. 9 is a property of that distribution's fitted quantile function, not an independent confirmation of the generative assumption λ_j = j. The 'validation' therefore re-describes the fitted tail rather than testing the model against new information; the paper itself concedes the procedure 'has limited practical significance.'

full rationale

The core mathematical derivation in Sec. III A 1 is not circular: it attempts to derive the tail exponent α = β − 1/2 from stated assumptions (λ_j = j, A_j ∼ j^{-β}) using an asymptotic saddle-point argument. Even if that argument contains a summation error (a correctness concern, not a circularity concern), the claim is not equivalent to its input by construction. There is no load-bearing self-citation, no imported uniqueness theorem, and no ansatz smuggled in via citation. The circularity is confined to the empirical validation in Sec. IV B. There, the fitted global exponent α is used to define region areas via β = α + 1/2, so the subsequent 'recovery' of a power-law area distribution is a definitional tautology. Moreover, sorting balls by count before measuring region intensities means the 'recovered' linear intensity growth is an order-statistic artifact of the same fitted distribution, not an independent prediction. The paper's own admission that the procedure 'has limited practical significance' supports treating this as partial, validation-level circularity rather than a collapse of the whole theoretical framework. The mathematical derivation, while possibly wrong, has independent content and is not itself circular.

Assumptions & free parameters 7 free parameters · 4 assumptions · 1 invented entities

The central derivation depends on an ad hoc power-law coupling between region area and Poisson intensity, plus the assumption of local Poisson homogeneity. The empirical validation adds free parameters (mixture weights, region counts, DBSCAN settings) that are fit or chosen post hoc. No invented physical entities are introduced beyond the statistical construction of non-contiguous intensity-sharing regions.

free parameters (7)
  • ball radius r = 50 m = 50 m
    Chosen as a trade-off between resolution and coverage (Sec II B); affects all counts and is not varied.
  • Poisson intensity index lambda_j = j = j
    Ad hoc integer indexing of intensities in Sec III A 1; any increasing sequence would make the calculation work.
  • area scaling exponent beta = 2.5 in synthetic; data-derived via alpha + 1/2
    Controls the power-law decay of region areas; assumed A_j ~ j^{-beta} without derivation from data.
  • DBSCAN parameters (epsilon, minPts) = 30 combinations
    epsilon varied 50 m to 7 km, minPts 50 to 3; representative clusters selected for display without a systematic criterion.
  • number of regions in re-clustering = not specified
    Sec IV B says 'predefine the number of regions' but the number is never reported.
  • mixture weights and intensities (pi_cj, lambda_cj) = fit to data
    Hybrid model components are fit to empirical counts; no predictive test is performed.
  • xmin for power-law fits = not reported
    The Clauset method requires selection of xmin; the paper reports exponents and p-values but not xmin values.
assumptions (4)
  • domain assumption Within each region, POI counts are i.i.d. Poisson with fixed rate lambda_j (homogeneous Poisson point process).
    Core modeling assumption in Sec III A, Level 2; tested only on selected DBSCAN clusters, and DBSCAN density thresholds can force apparent homogeneity.
  • domain assumption Counts in different balls are independent samples from the global mixture.
    Used for KS tests and the mixture aggregation, but overlapping 50 m balls create strong spatial correlation, violating independence.
  • ad hoc to paper Areas and intensities couple as A_j ~ j^{-beta} and lambda_j = j.
    Assumed in Sec III A 1 to make the sum analytically tractable; the paper acknowledges real clusters do not have such clean relationships.
  • standard math Robbins' Stirling bounds and the saddle-point approximation are valid in the joint limit C, k -> infinity.
    Used in Eq. (4); the saddle-point step is where the exponent error enters.
invented entities (1)
  • Regions as non-contiguous unions of clusters sharing a common Poisson intensity
    purpose: Allows areas with similar POI density to be spatially disjoint while sharing one rate, enabling the global-mixture calculation.
    A modeling construct introduced in Sec III A; there is no direct empirical evidence that such regions are the true generating units of POI counts.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scale-free Points-of-Interest Distribution in a City Emerging from Homogeneous Poissonian-point Processes." pith.science (2026). https://pith.science/paper/GVQ3YVNY

@misc{pith2026250901699,
  author       = {Pith},
  title        = {Pith review of: Scale-free Points-of-Interest Distribution in a City Emerging from Homogeneous Poissonian-point Processes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GVQ3YVNY}},
  note         = {Machine review of arXiv:2509.01699}
}
read the original abstract

Urban systems often exhibit scale-invariant properties, with power-law distributions observed in various spatial and temporal patterns of human behavior. A prominent example is the distribution of commercial activities and other Points of Interest (POIs) across cities. However, the mechanisms by which such heavy-tailed behaviors emerge from local urban dynamics remain poorly understood. In this work, we demonstrate that global inhomogeneity in the spatial distribution of POIs can arise from the aggregation of locally homogeneous processes. Using Foursquare data from the city of Bologna, we show that POI distributions exhibit clear power-law scaling when analyzed at city scale. We develop a theoretical framework in which this behavior naturally emerges from spatial clusters defined by shared intensity levels across disjoint areas, rather than spatial contiguity. By analytically and empirically linking these local processes to the observed global distribution, we provide a generative explanation for the emergence of scale-free patterns in urban commercial structure. To further relax the assumptions underlying the purely spatial model, and to account for the empirical observation that areas with similar activity intensity can be spatially disjoint, we introduce a hybrid hierarchical approach that combines spatial clustering with statistical heterogeneity across regions of comparable density, modeled via Poisson mixtures. This enables us to capture real-world deviations from local regularity while preserving interpretability. Our findings highlight a key insight: complex global phenomena in cities can arise from the spatial superposition of simple, locally uniform dynamics. This connection between micro-level homogeneity and macro-scale complexity offers new tools for interpreting, modeling, and classifying urban space.

Figures

Figures reproduced from arXiv: 2509.01699 by the authors.

Figure 1
Figure 1. FIG. 1. Number of POIs per Foursquare category and district. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Visualization of Bologna’s districts, displayed with [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Entire dataset plotted on a logarithmic scale. (a) [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Density-based clustering of POIs in Bologna, strati [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Histograms of POI counts per ball for selected clus [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Synthetic composite surface generated by partitioning [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Distribution of the number of points per unit ball [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 9
Figure 9. Figure 9: FIG. 9. Left: Unitary balls on the map colored by regions, [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 10
Figure 10. Figure 10: FIG. 10. Toy-model illustration of the ball covering procedure. The urban surface is first covered by a random Boolean model of [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: FIG. 11. Pie charts showing the proportion of Bologna’s total urban surface occupied by POI clusters, identified through [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: FIG. 12. Pie charts showing the proportion of Bologna’s total urban surface occupied by POI clusters, identified through [PITH_FULL_IMAGE:figures/full_fig_p018_12.png]
Figure 13
Figure 13. Figure 13: FIG. 13. Zoomed view of non-dominant POI clusters’ surface area per category. [PITH_FULL_IMAGE:figures/full_fig_p019_13.png]
Figure 14
Figure 14. Figure 14: FIG. 14. Zoomed view of non-dominant POI clusters’ surface area per category. [PITH_FULL_IMAGE:figures/full_fig_p020_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 38 canonical work pages

  1. [1]

    Let λj = j for j ∈ N, corresponding to discrete Poisson components with increasing rate

    Local Poisson Distributions Leading to Global Power-Law Decay The goal is to show that local Poisson distributions can give rise to a global power-law decay. Let λj = j for j ∈ N, corresponding to discrete Poisson components with increasing rate. This choice reflects the fact that the number of POIs per ball is integer-valued, making it natural to index t...

    work page

  2. [2]

    This interpretation serves both to clarify the logic of the model and to enable direct comparison with the alternative mixture-based approach introduced in the next section

    Three-Level Interpretation of the Local Poisson Model To summarize the modeling assumptions underlying the local Poisson framework, we introduce a structured three-level interpretation that connects the local genera- tion of POIs with the emergence of global distributional patterns. This interpretation serves both to clarify the logic of the model and to ...

    work page

  3. [3]

    Barthelemy, The Structure and Dynamics of Cities: Urban Data Analysis and Theoretical Modeling (Cam- bridge University Press, Cambridge, 2016)

    M. Barthelemy, The Structure and Dynamics of Cities: Urban Data Analysis and Theoretical Modeling (Cam- bridge University Press, Cambridge, 2016)

    work page 2016

  4. [4]

    L. M. A. Bettencourt, The origins of scaling in cities, Science 340, 1438 (2013)

    work page 2013

  5. [5]

    Kaufmann, L

    T. Kaufmann, L. Radaelli, L. M. A. Bettencourt, et al. , Scaling of urban amenities: generative statistics and im- plications for urban planning, EPJ Data Science 11, 50 (2022)

    work page 2022

  6. [6]

    Rybski, A

    D. Rybski, A. Garc ´ ıa Cant´ u Ros, and J. P. Kropp, Distance-weighted city growth, Phys. Rev. E 87, 042114 (2013)

    work page 2013

  7. [7]

    Huang and C

    X. Huang and C. Yost-Bremm, The power law within a metropolitan area, Cities 72, 201 (2018)

    work page 2018

  8. [8]

    G. K. Zipf, Human Behavior and the Principle of Least Effort (Addison-Wesley, Cambridge, MA, 1949) p. 573

    work page 1949

Show all 41 references
  1. [9]

    Verbavatz and M

    V. Verbavatz and M. Barthelemy, The growth equation of cities, Nature 587, 397 (2020)

    2020

  2. [10]

    Lemoy and G

    R. Lemoy and G. Caruso, Scaling laws in the spatial structure of urban road networks, Physica A: Statistical Mechanics and its Applications 474, 476 (2017)

    2017

  3. [11]

    H. A. Makse, S. Havlin, and H. E. Stanley, Modelling urban growth patterns, Nature 377, 608 (1995)

    1995

  4. [12]

    Bazzani, B

    A. Bazzani, B. Giorgini, S. Rambaldi, R. Gallotti, and L. Giovannini, Statistical laws in urban mobility from microscopic gps data in the area of florence, Journal of Statistical Mechanics: Theory and Experiment 2010, P05001 (2010)

    2010

  5. [13]

    Alessandretti, U

    L. Alessandretti, U. Aslak, and S. Lehmann, The scales of human mobility, Nature 587, 402 (2020)

    2020

  6. [14]

    Mizzi, A

    C. Mizzi, A. Baroncini, A. Fabbri, D. Micheli, A. Van- nelli, C. Criminisi, S. Jean, and A. Bazzani, Individ- ual mobility deep insight using mobile phones data, EPJ Data Science 12, 56 (2023)

    2023

  7. [15]

    Gravier and M

    J. Gravier and M. Barthelemy, A typology of activities over a century of urban growth, Nature Cities 1, 567 (2024)

    2024

  8. [16]

    R. D. Malmgren, D. B. Stouffer, A. S. L. O. Campanharo, and L. A. N. Amaral, Universality of human correspon- dence activity, Science 325, 1696 (2009)

    2009

  9. [17]

    R. D. Malmgren, D. B. Stouffer, A. E. Motter, and L. A. N. Amaral, A poissonian explanation for heavy tails in e-mail communication, Proceedings of the Na- tional Academy of Sciences 105, 18153 (2008)

    2008

  10. [18]

    Li and P

    Z.-W. Li and P. He, Data-based optimal bandwidth for kernel density estimation of statistical samples*, Com- munications in Theoretical Physics 70, 728 (2018)

    2018

  11. [19]

    Gallotti, G

    R. Gallotti, G. Bertagnolli, and M. De Domenico, Unrav- eling the hidden organisation of urban systems and their mobility flows, EPJ Data Science 10, 3 (2021)

    2021

  12. [20]

    Batty, The size, scale, and shape of cities, Science 319, 769 (2008)

    M. Batty, The size, scale, and shape of cities, Science 319, 769 (2008)

    2008

  13. [21]

    G. J. McLachlan and D. Peel, Finite mixture models , Probability and Statistics – Applied Probability and Statistics Section, Vol. 299 (Wiley, New York, 2000)

    2000

  14. [22]

    com/ (accessed on 15 January 2025)

    Foursquare, Available online: https://foursquare. com/ (accessed on 15 January 2025)

    2025

  15. [23]

    comune.bologna.it/amministrazione (2025), accessed on 19 June 2025

    Comune di Bologna, Amministrazione, https://www. comune.bologna.it/amministrazione (2025), accessed on 19 June 2025

    2025

  16. [24]

    OpenStreetMap contributors, Openstreetmap, https:// www.openstreetmap.org (2025), accessed: 2025-08-19

    2025

  17. [25]

    comune.bologna.it/ (2025), district boundaries dataset

    Comune di Bologna, Open data bologna, https://dati. comune.bologna.it/ (2025), district boundaries dataset. Accessed: 2025-08-19

    2025

  18. [26]

    Jost, Riemannian Geometry and Geometric Analysis , 7th ed., Universitext (Springer, 2017)

    J. Jost, Riemannian Geometry and Geometric Analysis , 7th ed., Universitext (Springer, 2017)

    2017

  19. [27]

    Pareto, Cours d’´ economie politique(F

    V. Pareto, Cours d’´ economie politique(F. Rouge, Lau- sanne, 1897)

  20. [28]

    Bak, How Nature Works: The Science of Self- Organized Criticality (Copernicus, New York, 1996)

    P. Bak, How Nature Works: The Science of Self- Organized Criticality (Copernicus, New York, 1996)

    1996

  21. [29]

    Barab´ asi and R

    A.-L. Barab´ asi and R. Albert, Emergence of scaling in random networks, Science 286, 509 (1999)

    1999

  22. [30]

    Clauset, C

    A. Clauset, C. R. Shalizi, and M. E. J. Newman, Power- law distributions in empirical data, SIAM Review51, 661 (2009)

    2009

  23. [31]

    R. D. Malmgren, D. B. Stouffer, A. S. L. O. Campan- haro, and L. A. N. Amaral, On universality in human correspondence activity, Science 325, 1696 (2009)

    2009

  24. [32]

    Gallotti, A

    R. Gallotti, A. Bazzani, S. Rambaldi, and M. Barthelemy, A stochastic model of randomly accelerated walkers for human mobility, Nature Commu- nications 7, 12600 (2016). 15

    2016

  25. [33]

    M. L. Goldstein, S. A. Morris, and G. G. Yen, Problems with fitting to the power-law distribution, The European Physical Journal B - Condensed Matter and Complex Systems 41, 255 (2004)

    2004

  26. [34]

    X. Ran, Y. Xi, Y. Lu, X. Wang, and Z. Lu, Comprehen- sive survey on hierarchical clustering algorithms and the recent developments, Artificial Intelligence Review 56, 8219 (2023)

    2023

  27. [35]

    Ester, H.-P

    M. Ester, H.-P. Kriegel, J. Sander, and X. Xu, A density- based algorithm for discovering clusters in large spatial databases with noise, in Proceedings of the Second Inter- national Conference on Knowledge Discovery and Data Mining (KDD) (AAAI Press, Portland, OR, USA, 1996) ...

    1996

  28. [36]

    Schubert, J

    E. Schubert, J. Sander, M. Ester, H.-P. Kriegel, and X. Xu, DBSCAN revisited, revisited: Why and how you should (still) use DBSCAN, ACM Transactions on Database Systems (TODS) 42, 19:1 (2017)

    2017

  29. [37]

    A. E. Ezugwu, A. M. Ikotun, O. O. Oyelade, L. Abuali- gah, J. O. Agushaka, C. I. Eke, and A. A. Akinyelu, A comprehensive survey of clustering algorithms: State-of- the-art machine learning applications, taxonomy, chal- lenges, and future research prospects, Engineering Ap- pl...

    2022

  30. [38]

    Stoyan, W

    D. Stoyan, W. S. Kendall, and J. Mecke, Stochastic Ge- ometry and Its Applications, 2nd ed. (John Wiley & Sons, Chichester, UK, 1995)

    1995

  31. [39]

    Hackl and B

    J. Hackl and B. T. Adey, Generation of spatially em- bedded random networks to model complex transporta- tion networks, in 14th International Probabilistic Work- shop, edited by R. Caspeele, L. Taerwe, and D. Proske (Springer International Publishing, Cham, 2017) pp. 217–230

    2017

  32. [40]

    Robbins, A remark on stirling’s formula, The Ameri- can Mathematical Monthly 62, 26 (1955)

    H. Robbins, A remark on stirling’s formula, The Ameri- can Mathematical Monthly 62, 26 (1955)

    1955

  33. [41]

    G. E. Willmot and X. S. Lin, Mixed poisson distributions, in Lundberg Approximations for Compound Distributions with Insurance Applications (Springer New York, New York, NY, 2001) pp. 37–49. APPENDIX: ADDITIONAL FIGURES 16 FIG. 10. Toy-model illustration of the ball covering p...

    2001

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.