OpenAI preprint, September 2026 · §6. Overlap conventions and dimension of the attractor
§6 (setting): the attractor and similarity dimension:Finite nonempty families of strict nonzero similarities of the line
Coding the attractor and defining similarity dimension.
Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its
declaration both in Lean and in words.
The Setting
Let be a finite nonempty alphabet and let be a family of affine similarities of the real line satisfying (1) An infinite address is a sequence ; a finite word is a string of symbols in . Write for the empty word and for the word obtained by placing before .
The Claim
Definition 1. The coding point of an infinite address is defined by (2), with an empty product equal to :(2)✦The coding attractor is the range of the coding map, as in (3): (3)✦The affine word coefficients are defined by the recursion (4); the associated map is , with words composed in address order, :(4)✦
Theorem 1. For every finite nonempty family satisfying (1), there is a unique nonnegative real number satisfying (5); this number is the similarity dimension of .✦(5)✦
In the paper, the result reads: is the compact attractor, equivalently the set of all coding limits, and is the unique nonnegative solution of , which exists for every finite nonempty family of strict nonzero similarities and equals zero for a single symbol.
What Is Checked
The coding constructions and the similarity dimension theorem are stated over finite nonempty alphabets and real affine families satisfying (1).