Basin Adjacency Graphs: A New Approach to Characterizing Basins of Attraction

arXiv:2611.15626 · nlin.CD · Submitted 2026-11-11 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Nonlinear Sciences. The day in nonlinear science: chaos, patterns, networks and adaptation.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Basin Adjacency Graphs".

Mira: Here's the gist of the paper. Fractal basins of attraction are commonly characterized through resolution-dependent measurements, such as box-counting methods, uncertainty exponents, and basin entropy Here we introduce a different approach based on the adjacency relations between connected components of the basins, without retaining their detailed spatial geometry.

Kai: The paper proposes a new method for characterizing basins of attraction by representing a finite-resolution basin diagram as a graph where each connected basin region becomes a vertex and neighboring regions are linked by edges.

Mira: This representation reveals how basin regions are organized and how additional components emerge as the spatial resolution increases.

Kai: The construction is inspired by region-adjacency graphs used in image analysis, where each region is represented by a vertex and neighboring regions are connected by edges.

Paper summary: Kai: So to wrap up this paper "Basin Adjacency Graphs: A New Approach to Characterizing Basins of Attraction" it introduces a method where we represent basins as graphs based on component adjacency instead of focusing on the detailed spatial geometry <ref:2611.0000>

Mira: It claims that the vertex-scaling exponent alpha, which describes how many vertices there are for resolution N, is linked to the Korčák exponent and can match the box-counting dimension of the basin boundary under specific assumptions <ref:2611.15626#pg3>

Lev: This means that this scaling isn't just about a single boundary but it points to how components proliferate within the underlying two dimensional domain <ref:2611.0000>

Kai: The paper also suggests that this graph structure offers complementary information because the degree distribution contains details about component organization that you wouldn't get just by looking at size distributions <ref:2611.0000>

Mira: But the authors are upfront about a limitation, which is that fractal boundaries separating a finite number of connected regions can result in graphs with a fixed number of vertices regardless of the boundary's actual fractal dimension <ref:2611.15626#pg2>

Lev: So while this graph scaling is promising for characterizing proliferation we still can't determine local properties like the exact Wada condition from component adjacency alone <ref:200#pg7>

Kai: That means this method gives us information on connectivity beyond just how big the basins are <ref:2611.0000>

Mira: It’s a compact way to represent these spatial maps that keeps the organization of components in mind <ref:2611.0000>

Lev: The degree distribution, which shows how many neighbors a region has, gives us clues about the overall layout of those basins <ref:2617B#pg8>

Kai: So what does this mean for the bigger picture when we think about these kinds of complex maps <ref:2611.0000>

Mira: The paper "Basin Adjacency Graphs: A New Approach to Characterizing Basins of Attraction" is fundamentally about changing how we look at basins of attraction by focusing on the way different regions touch each other <ref:2611.0000>

Lev: Exactly, it's not just looking at the shape of the attractor itself but how those different regions are connected to one another <ref:2611.0000>

Kai: The main technical thing they do is taking a finite resolution basin diagram and turning it into a graph where each region is a vertex and edges link adjacent components belonging to different basins <ref:2617B#pg5>

Mira: And they use that graph structure to see how the number of vertices scales with the spatial resolution N <ref:200#pg8>

Lev: They link that scaling exponent alpha to things like the Korčák exponent and even can match the box-counting dimension of a fractal boundary when its fractal complexity arises from the proliferation of components <ref:2611.15626#pg3>

Kai: But they also found that just looking at the average number of neighbors doesn't tell you if two components actually share a boundary <ref:200#pg7>

Mira: That means this method gives us information on connectivity beyond just how big the basins are and how many neighbors a region has <ref:2611.0000>

Kai: It suggests that this approach could be applied to characterizing categorical spatial maps beyond nonlinear dynamics <ref:2611.0000>

Mira: The whole point is gaining a compact representation by focusing on adjacency relations while retaining the scaling properties of component organization <ref:2611.15626#pg3>

Conclusion: Kai: So, we're looking at this paper "Basin Adjacency Graphs: A New Approach to Characterizing Basins of Attraction" and what it does <ref:2611.0000>

Mira: It basically says instead of measuring the shape of the basin itself, we should look at how all the different basin regions touch each other <ref:2617B#pg5>

Lev: The authors build these graphs by taking a resolution diagram, making each region a vertex and linking neighbors with edges <ref:2617B#pg5>

Kai: They then measure how many vertices there are as the spatial resolution N goes up <ref:200#pg8>

Mira: And that scaling exponent alpha they find is related to the Korčák exponent and even matches the fractal dimension of a boundary when it comes from component proliferation <ref:2611.15626#pg3>

Lev: But there's a catch, they note that just counting how many neighbors a region has doesn't prove two components actually share a boundary <ref:200#pg7>

Kai: It means this method gives us information about the connectivity of different regions beyond just how big the basins are <ref:2611.0000>

Mira: It’s a compact way to represent these spatial maps by focusing on those adjacency relations while keeping track of how the components are organized <ref:2617B#pg8>

Lev: The degree distribution, which shows how many neighbors a region has, gives us clues about the overall layout of those basins <ref:2617B#pg8>

Kai: So what does this mean for the bigger picture when we think about these kinds of complex maps? <ref:2611.0000>

Alexandre Wagemakers, Aryan Srivastava, Miguel A. F. Sanjuáná

Nonlinear Dynamics, Chaos and Complex Systems Group, Universidad Rey Juan Carlos · School of Mathematical Sciences, UM-DAE Centre for Excellence in Basic Sciences (CEBS), University of Mumbai · Royal Academy of Sciences of Spain

nlin.CD

Submitted: 2026-11-11

Updated: 2026-11-11

Code: https://github.com/awage/AnalysisBasinsNets

License: http://creativecommons.org/licenses/by-sa/4.0/

Importance score: 77/100

The gist: Fractal basins of attraction are commonly characterized through resolution-dependent measurements, such as box-counting methods, uncertainty exponents…

Key concepts

Region-Adjacency Graph (BAG)
This method represents a finite-resolution basin diagram as a graph where each connected region is a vertex, and edges connect neighboring regions. It focuses on the adjacency and connectivity of basins rather than their exact spatial geometry, allowing researchers to study how components proliferate with increasing resolution.
Vertex Scaling Exponent ($\alpha$)
This exponent describes how the number of graph vertices (basin regions) scales with spatial resolution ($N$), following a power law $N^{\alpha}$. This scaling is related to the Korčák exponent and can coincide with the box-counting dimension of basin boundaries when components proliferate, suggesting a link between connectivity structure and fractal geometry.
Mean Basin-Neighbor Count ($\bar{w}$)
This statistic measures how many neighbors each basin has on average. For systems with the Wada property, this count can approach its maximum value as resolution increases. However, it does not prove the Wada property alone because adjacent components do not necessarily share a boundary.
Degree Distribution
The degree distribution analyzes how many neighbors different basin components have. Large connected regions often have high degrees because they are adjacent to many smaller ones, providing information about the spatial arrangement of components that is independent of just their size.

Terminology

Summary

Fractal basins of attraction are commonly characterized through resolution-dependent measurements, such as box-counting methods, uncertainty exponents, and basin entropy Here we introduce a different approach based on the adjacency relations between connected components of the basins, without retaining their detailed spatial geometry. p. 0

Where the paper says it

Sanjuán1, 3 1Nonlinear Dynamics, Chaos and Complex Systems Group, Departamento de Física, Universidad Rey Juan Carlos, Tulipán s/n, 28933 Móstoles, Madrid, Spain 2School of Mathematical Sciences, UM-DAE Centre for Excellence in Basic Sciences (CEBS), University of Mumbai, Vidyanagari, Mumbai 400098, India 3Royal Academy of Sciences of Spain, Valverde 22, 28004 Madrid, Spain Fractal basins of attraction are commonly characterized through resolution-dependent measurements, such as box-counting methods, uncertainty exponents, and basin entropy. Here we introduce a different approach based on the adjacency relations between connected components of the basins, without retaining their detailed spatial geometry.

Introduction to the Approach

The paper proposes a new method for characterizing basins of attraction by representing a finite-resolution basin diagram as a graph where each connected basin region becomes a vertex and neighboring regions are linked by edges. This representation reveals how basin regions are organized and how additional components emerge as the spatial resolution increases. p. 0

Where the paper says it

We describe these patterns using a graph in which each connected basin region becomes a vertex and neighboring regions are linked by edges. This representation reveals how basin regions are organized and how additional components emerge as the spatial resolution increases.

The construction is inspired by region-adjacency graphs used in image analysis, where each region is represented by a vertex and neighboring regions are connected by edges. p. 0
Where the paper says it

Rather than describing their detailed spatial geometry, we represent a finiteresolution basin diagram by a graph. The construction is inspired by region-adjacency graphs used in image analysis [8], in which each region is represented by a vertex and neighboring regions are connected by edges.

This approach differs from recurrence networks and related constructions because it describes the adjacency and connectivity of basin components rather than the structure of the attractors themselves. p. 0
Where the paper says it

Recurrence networks and related constructions encode time series as networks whose properties reflect the geometry of attractors [9, 10], while the qualitative behavior of a map can be represented through a graph of its chain-recurrent classes [11]. The BAG differs from these approaches in that it describes the adjacency and connectivity of basin components rather than the structure of the attractors themselves.

Vertex Scaling and Fractal Dimensions

The study investigates how the number of graph vertices scales with spatial resolution N, following a power law N α in the systems considered. Under suitable assumptions on the size distribution and geometry of the components, this vertex-scaling exponent α is related to the Korčák exponent of the component-size distribution and may coincide with the box-counting dimension of the basin boundary when its fractal structure arises from the proliferation of components. p. 1

Where the paper says it

2 nally introduced in the study of island sizes [12] and subsequently used to characterize the spatial patchiness of landscapes [13, 14]. Under suitable assumptions on the size distribution and geometry of the components, the vertex-scaling exponent is related to the Korčák exponent and may coincide with the box-counting dimension of the basin boundary when its fractal complexity arises from the proliferation of components.

For an idealized Sierpiński construction, it is shown that for large m, V(ε) ≃ 3(m/2) = 3(m/2) = ε(-df), where df is the Hausdorff dimension of the limiting Sierpiński triangle. p. 2
Where the paper says it

Expressing the vertex count in terms of this scale, we obtain for large m |V (ε)| ≃ 3 2 3 m = 3 2 ε − log 3/ log 2 = 3 2 ε −df , (5) where df = log 3/ log 2 ≈ 1. 585 is the Hausdorff dimension of the limiting Sierpiński triangle.

This illustrates how component proliferation can yield an exponent equal to a geometrical fractal dimension. p. 2
Where the paper says it

This idealized convention differs from the eight-neighbor pixel connectivity used for the numerical BAGs. The Sierpiński triangle thus illustrates how component proliferation can yield an exponent equal to a geometrical fractal dimension (Fig.

Numerical Results and Comparisons

The results for Newton’s method show a vertex-scaling exponent α = 1.424, which is consistent with independent estimates of the fractal boundary dimension. p. 3

Where the paper says it

A log-log fit over 100 ≤ N ≤ 2000 gives α = 1. 424, consistent with independent estimates of the fractal boundary dimension.

The Korčák estimate 2K at N = 2000 gives 1. p. 3
Where the paper says it

1 ≤ h ≤ 1. 9: boxcounting estimate db and vertex-scaling exponent α, obtained using resolutions N = 250–2000, together with the single-grid Korčák estimate 2K at N = 2000, plotted against the boxcounting dimension dF reported by Saupe [24].

38, differing by less than 0. p. 3
Where the paper says it

At the highest resolution, N = 2000, this procedure gives 2K = 1. 38, differing by less than 0.

07 from both estimates of α and from the independently determined boundary dimension. p. 3
Where the paper says it

38, differing by less than 0. 07 from both estimates of α and from the independently determined boundary dimension.

For the relaxed Newton family, the vertex-scaling exponent α also shows good agreement with the box-counting dimension dF within 0. p. 3
Where the paper says it

The relaxed Newton family To further assess the vertex-scaling approach, we consider the relaxed Newton family Nh(z) = z−h p(z)/p′ (z), with p(z) = z 3 + (λ − 1)z − λ and λ = 1. For this family, Saupe computed the box-counting dimension dF (h) of the Julia set over the parameter range 0.

025 for h ≥ 0. p. 3
Where the paper says it

The box-counting estimate db agrees with dF within 0. 025 throughout the parameter range.

5.

Adjacency Descriptors and Limitations

The mean basin-neighbor count w¯ provides a simple measure of adjacency relations, and for systems exhibiting the Wada property, w¯ may approach its maximum value NA − 1 as the spatial resolution increases. However, this quantity does not by itself establish the Wada property because mutually adjacent components need not share the same boundary. p. 4

Where the paper says it

For basin configurations exhibiting the Wada property, w¯ may approach its maximum value NA − 1 as the spatial resolution increases. However, this condition is not sufficient to establish the Wada property, since mutually adjacent components need not share the same boundary.

A key limitation is that fractal boundaries separating a finite number of connected regions may produce graphs with a fixed number of vertices, independently of the geometrical complexity of the boundary. p. 5
Where the paper says it

The method has several limitations. First, fractal boundaries separating a finite number of connected regions, such as quasicircle Julia sets, may produce graphs with a fixed number of vertices, independently of the geometrical complexity of the boundary.

Conclusion and Future Directions

The BAG provides a compact representation by retaining adjacency relations, and its scaling properties originate from the spatial organization and proliferation of basin components in the underlying two-dimensional domain. The method offers complementary information about the connectivity of different regions beyond their size distributions. p. 5

Where the paper says it

Finally, the Korčák exponent has also been used to characterize the patchiness of landscapes and habitat fragmentation [13, 14]. The BAG could therefore offer a complementary description of categorical spatial maps, providing information about the connectivity and adjacency of different regions beyond their size distributions.

The degree distribution contains information about the organization of basin components that is not determined by the component-area distribution alone. p. 5
Where the paper says it

The number of neighboring components depends not only on their sizes but also on their spatial arrangement and mutual contacts. Thus, the degree distribution contains information about the organization of the basin components that is not determined by the componentarea distribution alone.

The approach suggests potential applications outside nonlinear dynamics for characterizing categorical spatial maps. p. 5
Where the paper says it

The BAG could therefore offer a complementary description of categorical spatial maps, providing information about the connectivity and adjacency of different regions beyond their size distributions. This suggests potential applications of the approach outside nonlinear dynamics.

TABLE I

System NA db α 2K Wada w¯

|---|---|---|---|---|---|

Newton z cubed - 1 3 1.443 1.447 (Table I)

--- Page 7 ---

The connected components of each basin are identified using eight-neighbor pixel connectivity, implemented through a union–find algorithm. The resulting basin adjacency graphs are constructed and analyzed using Graphs. p. 6

Where the paper says it

The connected components of each basin are identified using eight-neighbor pixel connectivity, implemented through a union–find algorithm. The resulting basin adjacency graphs are constructed and analyzed using Graphs.

jl. The vertex-scaling exponent α is estimated from a least-squares fit of log V against log N. p. 6
Where the paper says it

The resulting basin adjacency graphs are constructed and analyzed using Graphs. The vertex-scaling exponent α is estimated from a least-squares fit of log |V | against log N.

As a reference for the boundary dimension, we use a boxcounting estimate db, obtained from the scaling with N of the number of boundary pixels, defined as pixels having at least one eight-neighbor belonging to a different basin. p. 6
Where the paper says it

The vertex-scaling exponent α is estimated from a least-squares fit of log |V | against log N. As a reference for the boundary dimension, we use a boxcounting estimate db, obtained from the scaling with N of the number of boundary pixels, defined as pixels having at least one eight-neighbor belonging to a different basin.

The component-area distribution is also analyzed on the finest grid to estimate the Korčák exponent K, following the fitting procedure of Imre et al. p. 6
Where the paper says it

This corresponds to box counting with box size ε = 1/N. The component-area distribution is also analyzed on the finest grid to estimate the Korčák exponent K, following the fitting procedure of Imre et al.

[14]. The fit uses component areas between 10 pixels and 10−3 of the total image area. p. 6
Where the paper says it

The component-area distribution is also analyzed on the finest grid to estimate the Korčák exponent K, following the fitting procedure of Imre et al. The fit uses component areas between 10 pixels and 10−3 of the total image area.

For comparison, we also consider independent estimates of the boundary dimension obtained from the uncertainty exponent [3], as well as the merging-method criterion for identifying Wada basin boundaries. p. 6
Where the paper says it

The fit uses component areas between 10 pixels and 10−3 of the total image area. For comparison, we also consider independent estimates of the boundary dimension obtained from the uncertainty exponent [3], as well as the merging-method criterion for identifying Wada basin boundaries [28].

The adjacency statistics extracted from the BAG are interpreted separately from these independent tests, since component adjacency alone does not establish the Wada property. p. 6
Where the paper says it

For comparison, we also consider independent estimates of the boundary dimension obtained from the uncertainty exponent [3], as well as the merging-method criterion for identifying Wada basin boundaries [28]. The adjacency statistics extracted from the BAG are interpreted separately from these independent tests, since component adjacency alone does not establish the Wada property.

The code that reproduces all figures and tables is available at https://github. p. 6
Where the paper says it

The adjacency statistics extracted from the BAG are interpreted separately from these independent tests, since component adjacency alone does not establish the Wada property. The code that reproduces all figures and tables is available at https://github.

com/awage/AnalysisBasinsNets. p. 6
Where the paper says it

The code that reproduces all figures and tables is available at https://github. com/awage/ AnalysisBasinsNets.

The work was supported by the Spanish State Research Agency (AEI) and the European Regional Development Fund (ERDF, European Union) under Project No. p. 6
Where the paper says it

com/awage/ AnalysisBasinsNets. ACKNOWLEDGMENTS This work was supported by the Spanish State Research Agency (AEI) and the European Regional Development Fund (ERDF, European Union) under Project No.

PID2023-148160NB-I00. p. 6
Where the paper says it

ACKNOWLEDGMENTS This work was supported by the Spanish State Research Agency (AEI) and the European Regional Development Fund (ERDF, European Union) under Project No. PID2023-148160NB-I00 (MCIN/AEI/10.

--- Page 8 ---

The degree distribution provides additional information about the adjacency structure. Small components often have relatively few neighbors, whereas large connected components may be adjacent to many smaller ones, producing high-degree vertices. p. 5

Where the paper says it

The degree distribution provides additional information about the adjacency structure. Small components often have relatively few neighbors, whereas large connected components may be adjacent to many smaller ones, producing high-degree vertices.

In the systems considered, the mean degree remains between approximately 4 and 6 for Newton’s method, the magnetic pendulum, and the Feudel map. p. 5
Where the paper says it

Small components often have relatively few neighbors, whereas large connected components may be adjacent to many smaller ones, producing high-degree vertices. In the systems considered, the mean degree remains between approximately 4 and 6 for Newton’s method, the magnetic pendulum, and the Feudel map, despite the presence of vertices with much larger degrees.

Although the degree distribution may exhibit a broad or approximately power-law tail, its scaling cannot be inferred directly from the Korčák law. p. 5
Where the paper says it

In the systems considered, the mean degree remains between approximately 4 and 6 for Newton’s method, the magnetic pendulum, and the Feudel map, despite the presence of vertices with much larger degrees. Although the degree distribution may exhibit a broad or approximately power-law tail, its scaling cannot be inferred directly from the Korčák law.

The number of neighboring components depends not only on their sizes but also on their spatial arrangement and mutual contacts. p. 5
Where the paper says it

Although the degree distribution may exhibit a broad or approximately power-law tail, its scaling cannot be inferred directly from the Korčák law. The number of neighboring components depends not only on their sizes but also on their spatial arrangement and mutual contacts.

Exploring this relationship more systematically could provide an interesting direction for future work. p. 5
Where the paper says it

Thus, the degree distribution contains information about the organization of the basin components that is not determined by the componentarea distribution alone. Exploring this relationship more systematically could provide an interesting direction for future work.

Finally, the Korčák exponent has also been used to characterize the patchiness of landscapes and habitat fragmentation [13, 14]. p. 5
Where the paper says it

Exploring this relationship more systematically could provide an interesting direction for future work. Finally, the Korčák exponent has also been used to characterize the patchiness of landscapes and habitat fragmentation [13, 14].

The BAG could therefore offer a complementary description of categorical spatial maps, providing information about the connectivity and adjacency of different regions beyond their size distributions. p. 5
Where the paper says it

Finally, the Korčák exponent has also been used to characterize the patchiness of landscapes and habitat fragmentation [13, 14]. The BAG could therefore offer a complementary description of categorical spatial maps, providing information about the connectivity and adjacency of different regions beyond their size distributions.

--- Page 9 ---

The method has several limitations. First, fractal boundaries separating a finite number of connected regions, such as quasicircle Julia sets, may produce graphs with a fixed number of vertices, independently of the geometrical complexity of the boundary. p. 5

Where the paper says it

The method has several limitations. First, fractal boundaries separating a finite number of connected regions, such as quasicircle Julia sets, may produce graphs with a fixed number of vertices, independently of the geometrical complexity of the boundary.

Second, estimating the vertex-scaling exponent requires basin diagrams at several spatial resolutions, which may be difficult when only a single experimental image is available. p. 5
Where the paper says it

In these cases, the BAG does not capture the roughness of an individual boundary, and complementary methods, such as the uncertainty exponent or basin entropy [3, 5, 7], remain necessary. Second, estimating the vertex-scaling exponent requires basin diagrams at several spatial resolutions, which may be difficult when only a single experimental image is available.

Third, the graph retains adjacency information but not the spatial arrangement of the interfaces between components. p. 5
Where the paper says it

Although coarser representations can sometimes be obtained from such an image, the accessible scaling range is necessarily limited by its original resolution. Third, the graph retains adjacency information but not the spatial arrangement of the interfaces between components.

Consequently, local properties of the basin boundaries, including the pointwise Wada condition, cannot be determined from component adjacency alone. p. 5
Where the paper says it

Third, the graph retains adjacency information but not the spatial arrangement of the interfaces between components. Consequently, local properties of the basin boundaries, including the pointwise Wada condition, cannot be determined from component adjacency alone.

--- Page 9 ---

The paper is 2611.15626. Every sentence must end with a reference token The paper is 2611.0000.

--- Page 9 ---

The paper is 2611.0000.

--- Page 9 ---

The paper is 2619A The paper is 2619A

--- Page 7 ---

The paper is 2617B

--- Page 8 ---

The paper is 2000

--- Page 8 ---

The paper is N = 200 (a) Basins, N = 200 log V

--- Page 7 ---

The paper is αu = 1.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Basin Adjacency Graph Analysis for Fractal Complexity: The improved system can characterize basin structures by examining how the number of vertices in a finite-resolution basin diagram scales with spatial resolution N, using the exponent alpha as a proxy for boundary fractal dimensions when component proliferation is present.

  2. Fractal Dimension Estimation: The system can estimate fractal boundary dimensions by relating the vertex-scaling exponent alpha to the Korčák exponent of the component-size distribution, specifically obtaining an estimate of 2K which agrees with independent estimates of the boundary dimension for systems like Newton's method, provided components are sufficiently compact.

  3. Adjacency Pattern Characterization: The system can provide adjacency-based diagnostics by calculating quantities like the mean basin-neighbor count w bar (wbar), which reveals characteristic patterns in Wada basin configurations, although this is noted as an adjacency-based descriptor rather than a sufficient test for the Wada property.

Abstract

Fractal basins of attraction are commonly characterized through resolution-dependent measurements, such as box-counting methods, uncertainty exponents, and basin entropy. Here we introduce a different approach based on the adjacency relations between connected components of the basins, without retaining their detailed spatial geometry. A finite-resolution basin diagram is represented by a graph with one vertex per connected component and edges linking adjacent components belonging to different basins. We investigate how the number of vertices scales with the spatial resolution N, following a power law N α in the systems considered. Under suitable assumptions, the exponent α is related to the Korčák exponent of the component-size distribution and can coincide with the box-counting dimension of the basin boundary when its fractal structure arises from the proliferation of components. The graph also provides adjacency-based diagnostics that may help characterize basin configurations exhibiting the Wada property. We illustrate the approach using idealized Sierpiński constructions, the Newton and relaxed Newton families, and three dynamical systems with previously established Wada basins.

Related papers