Correlation integral¶

The probability, estimated from state pairs, that two independently sampled points on a trajectory or invariant measure lie within a specified distance.

Version: v2 · 2026-09-19 · History
Domain-specific #: 3923
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomain: Nonlinear Time Series and Dynamical Systems → Mathematics

Core Idea¶

Its small-radius scaling estimates correlation dimension, but temporal dependence, finite samples, noise, embedding choices and edge effects require exclusion windows and a defensible scaling region. State vectors are reconstructed or observed, pairwise distances are compared with radius epsilon, qualifying pairs are counted and normalized, and the log slope of the resulting cumulative fraction is examined across scales. The abstraction is therefore identified by a declared carrier, a transformation or constraint over that carrier, and an invariant that tells an analyst whether the named structure is genuinely present.

How would you explain it like I'm…

Counting Close Pairs

Imagine a bunch of dots scattered on paper. Look at every possible pair of dots and ask: are these two closer than the width of your finger? The correlation integral is the share of all pairs that are that close. Scientists check how that share changes as they use smaller and smaller 'fingers.'

The Close-Pair Fraction

Imagine measuring something that changes over time, like the weather or a swinging pendulum, and turning each moment into a dot in a special picture. The correlation integral is the fraction of all pairs of dots that are closer together than some chosen distance. Scientists compute it for many distances and watch how it changes as the distance gets smaller — that pattern estimates how complicated the shape is (its correlation dimension). They have to be careful: dots from moments right next to each other in time are close just because little time passed, so those pairs are skipped. Noise and not having enough data can also fool the result.

Pairwise Proximity Fraction

The correlation integral is a tool from the study of dynamical systems and nonlinear time series. First you get state vectors — either measured directly or reconstructed from a single time series by embedding, which stacks delayed copies of the signal. Then, for a radius ε, you count the pairs of state vectors closer than ε (using a chosen distance measure) and divide by the number of pairs, giving the fraction C(ε). How C(ε) scales as ε shrinks — the slope on a log-log plot — estimates the correlation dimension. To make this trustworthy, you exclude pairs that are close only because they're close in time (a Theiler window), and you look for a range of ε where the slope is steady, since noise, too few samples, embedding choices, and edge effects distort the very small and very large scales.

The correlation integral C(epsilon) is the normalized count of pairs of state vectors, observed or reconstructed by delay embedding from a time series, whose distance under a chosen norm is less than epsilon. It estimates the probability that two points drawn from the system's invariant measure lie within epsilon of each other, and its small-radius scaling, the slope of log C(epsilon) against log epsilon, estimates the correlation dimension. Reliable use requires specifying the dynamical system or series, the embedding and norm, the range of radii, and finite-sample normalization. Temporally adjacent points must be excluded via a Theiler window, because their closeness reflects temporal dependence rather than geometry. The scaling region must be chosen defensibly between the noise floor at small radii and saturation at large radii, with edge effects and uncertainty reported.

Scope of Application¶

Correlation integral belongs to nonlinear time series and dynamical systems and is useful where the analyst can specify the typed nonlinear time series and dynamical systems carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, then evaluate the dynamical system or time series, state vectors and embedding, norm, radius, pair exclusions and Theiler window, finite-sample normalization, invariant-measure assumption, scaling interval, correlation dimension estimate, noise floor, saturation and uncertainty are explicit. The scope is broad within that domain but bounded by the need for the dynamical system or time series, state vectors and embedding, norm, radius, pair exclusions and Theiler window, finite-sample normalization, invariant-measure assumption, scaling interval, correlation dimension estimate, noise floor, saturation and uncertainty are explicit.

Clarity¶

The abstraction clarifies a crowded vocabulary by making the dynamical system or time series, state vectors and embedding, norm, radius, pair exclusions and Theiler window, finite-sample normalization, invariant-measure assumption, scaling interval, correlation dimension estimate, noise floor, saturation and uncertainty are explicit the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test.

Manages Complexity¶

Without the abstraction, an analyst must reason directly over many local details: the carrier roles, admissibility assumptions, competing conventions, derived invariants, boundary cases, and proof or validation obligations specific to Correlation integral. Correlation integral compresses them into the roles in the structural signature. That compression permits comparison across instances without erasing the variables that determine validity. It also exposes which details may be varied safely and which are constitutive.

Abstract Reasoning¶

Identify the carrier. State what the elements, states, objects, or observations are: the typed nonlinear time series and dynamical systems carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets. Reject examples whose alleged carrier belongs to a different problem. 2.

Knowledge Transfer¶

Knowledge transfers strongly among subfields of nonlinear time series and dynamical systems because they reuse the typed nonlinear time series and dynamical systems carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, State vectors are reconstructed or observed, pairwise distances are compared with radius epsilon, qualifying pairs are counted and normalized, and the log slope of the resulting cumulative fraction is examined across scales., and type the carrier, state every parameter and convention in the definition, test that the dynamical system or time series, state vectors and embedding, norm, radius, pair exclusions and Theiler window, finite-sample normalization, invariant-measure assumption, scaling interval, correlation dimension estimate, noise floor, saturation and uncertainty are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions¶

Current abstraction Correlation integral Domain-specific

Parents (1) — outgoing typed relations

Correlation integral is a kind of Measurement Prime

It remains its own entry because its identity is fixed by the dynamical system or time series, state vectors and embedding, norm, radius, pair exclusions and Theiler window, finite-sample normalization, invariant-measure assumption, scaling interval, correlation dimension estimate, noise floor, saturation and uncertainty.

Hierarchy path (1) — routes to 1 parentless root

Correlation integral → is a kind of Measurement

Neighborhood in Abstraction Space¶

Correlation integral sits in a crowded region of the domain-specific corpus (15^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Time-Series & Process-Variation Analysis (17 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25