Skip to content

Takens Formalization

A Lean 4 and Mathlib formalization of delay-coordinate reconstruction, from finite state spaces to compact manifolds.

Every selected declaration depends only on Lean's standard logical axioms. There are no sorrys, no custom axioms and no unproved infrastructure assumptions.

Read the Exposition Theorem Catalog


What is proved

The delay map of dynamics \(T : X \to X\) and an observation \(h : X \to \mathbb{R}\) is \(x \mapsto (h(x), h(Tx), \dots, h(T^{k-1}x))\), delayEmbedding in Lean.

Area Result Module
Manifolds Takens' theorem for generic pairs: on a compact \(d\)-manifold, the pairs \((T, h)\) of a \(C^2\) diffeomorphism and a \(C^2\) observation whose delay map with \(2d+1\) coordinates is a \(C^2\) embedding form an open dense subset of \(\mathrm{Diff}^2(M) \times C^2(M, \mathbb{R})\) GenericPairTakens
Manifolds Bounded-period nondegeneracy and observability density (a Kupka--Smale-type density lemma): the \(C^2\) diffeomorphisms \(T\) such that at every point of minimal period \(0 < p \le 4d\) the differential \(A = D(T^p)\) has \(A^m - 1\) invertible for \(1 \le m \le 4d\) and is observable are dense KupkaSmale, PatchPerturbation, PeriodicNull
Manifolds For an injective \(C^2\) map \(T\) with injective differentials whose points of period \(\le 4d\) are countably many and which is observable at points of period \(\le 2d\), the \(C^2\) observations whose delay map with \(2d+1\) coordinates is a \(C^2\) embedding form an open dense set in the \(C^2\) topology GenericObservation
Manifolds One finite family of smooth functions \(\varphi_q\) such that, for every such \(T\) and every \(C^2\) observation \(h\), the delay map of \(h + \sum_q a_q \varphi_q\) is a \(C^2\) embedding for Lebesgue-almost every \(a\) InterpolatingFamily, DelayPeriodic
Manifolds The weak \(C^n\) topology on \(C^n\) maps through chart derivatives; injective immersions of a compact manifold are stable under \(C^1\)-small perturbations JetTopology, EmbeddingStability
Manifolds The \(C^n\) topology on \(C^n\) diffeomorphisms through charts on source and target; closeness is preserved by composition and iteration; the pairs \((T, h)\) whose delay map is a \(C^2\) embedding are open in \(\mathrm{Diff}^2(M) \times C^2(M, \mathbb{R})\) WeakTopology, WeakComposition, GenericPair
Manifolds Differential of the delay map, immersion criterion, compact injective immersions are embeddings; the identity never immerses in dimension \(\ge 2\) SmoothDelay
Measure Sard's theorem for \(C^r\) maps between finite-dimensional spaces, \(r \ge \max\{1, \dim E - \dim F + 1\}\), with local and open-set versions Sard
Measure Almost every member of a finite-dimensional affine family avoids a lower-dimensional set, is an immersion, or separates points GenericFamily
Finite Injectivity iff orbit separation; the exact separating horizon; when some window separates, the sharp bound \(N-1\) on \(N \ge 1\) states, attained; a sound and complete decision procedure DelayWindow, TakensDiscrete
Finite Reconstruction of the dynamics on the delay image, a homeomorphism for compact spaces Reconstruction
Ordinal Ordinal patterns, invariance under strictly increasing and relabeling under strictly decreasing transformations, pattern counts and entropy bounds, what the code retains OrdinalTakens, OrdinalEntropy, OrdinalQuotient

Not formalized here: the Sauer--Yorke--Casdagli extension to fractal sets and prevalence; see Open Problems.


Documentation

Section What you'll find
Quickstart Build and verify the proofs
Proof Architecture How the modules fit together
Delay Embedding Injectivity and orbit separation
Coincidence Length Exact horizons on finite state spaces
Ordinal Compression What an ordinal code keeps
Smooth Embedding Differentials, immersions and generic observations
Sard's Theorem Critical values at finite regularity
Theorem Catalog The selected declarations, by module
Axiom Dashboard What the axiom records certify
Roadmap What comes next
Measured Data What these proofs do and do not say about measured time series