Delay-Coordinate Embedding — Takens' Theorem

Reconstructing attractors from scalar time series: x(t) → [x(t), x(t+τ), x(t+2τ), ...]

System

View

ModeEmbedding

Takens' Theorem

Given scalar x(t), form delay vectors:
y(t) = [x(t), x(t+τ), ..., x(t+(m-1)τ)]
For generic τ and m ≥ 2d_A+1
(d_A = attractor dimension),
the map is a diffeomorphic embedding.
False nearest neighbors → optimal m