Recursion & Self-Similarity (Fractals)

The unifying principle

Formally, an object $X$ is self-similar if it can be written as a union of scaled copies of itself:

$X = \bigcup_{i=1}^{N} f_i(X), \qquad f_i(x) = r_i x + b_i,\quad 0 < r_i < 1$

Each $f_i$ is a contraction (an iterated function system). Because the definition references $X$ on both sides, it is inherently recursive.

The "size" of such an object is captured by the fractal (Hausdorff) dimension. If $N$ copies each scaled by ratio $r$ reconstruct the whole, then

$N \cdot r^{D} = 1 \;\Longrightarrow\; D = \frac{\log N}{\log(1/r)}.$

This $D$ is often non-integer — the signature of a fractal. The same recursion, expressed as a process rather than a set, is a function invoking itself with a smaller argument until a base case halts it. Structure (fractals) and process (recursion) are two views of one fixed-point equation $X = F(X)$ .

How it shows up in each field

Maths — the fractal set

The Koch curve replaces each segment with 4 segments of length $\tfrac13$ : $D = \frac{\log 4}{\log 3} \approx 1.262.$ It is longer than a line ( $D=1$ ) but doesn't fill area ( $D=2$ ). The Mandelbrot set arises from iterating $z_{n+1} = z_n^2 + c$ — recursion in the complex plane.

Coding/CS — the self-calling process

A recursive procedure is the algorithmic incarnation of $X = F(X)$ :

def fib(n):
    if n < 2: return n        # base case = the f_i floor
    return fib(n-1) + fib(n-2)  # self-reference

Biology — branching that packs infinite surface into finite space

Lungs, blood vessels, and neurons branch recursively. The bronchial tree divides ~23 times, each branch a scaled copy of its parent, achieving $D \approx 2.97$ — almost space-filling — to cram ~ $70\,\text{m}^2$ of gas-exchange surface into the chest.

Stock-Market — statistical self-similarity in time

Price charts look statistically the same across timescales: a 1-minute chart resembles a daily chart. Mandelbrot modeled prices as fractional Brownian motion with a Hurst exponent $H$ : $P(t) \;\overset{d}{=}\; \lambda^{-H} P(\lambda t),$ so rescaling time by $\lambda$ and price by $\lambda^H$ leaves the distribution unchanged. Volatility clustering and fat tails follow from this, unlike the smooth Gaussian random walk ( $H = 0.5$ ).

Why this bridge matters

Dimension as complexity currency. The single number $D$ tells a mathematician about a curve's roughness, a physiologist about a lung's surface efficiency, and a trader about a market's roughness/persistence. One measurement transfers across all four.
Recursion tames the infinite. CS intuition — "solve the whole by solving a smaller copy, stop at the base case" — explains how a finite genome specifies an intricate branching organ (a recursive developmental rule) and how a finite fractal formula generates infinite detail.
Scale-invariance warns you. From coastlines and markets: "length" and "risk" have no scale-free absolute value. A market's fractal nature means variance can be undefined; the naïve Gaussian assumption underprices tail events — the 1987 crash is a $D$ -fingerprint, not a fluke.
Fixed-point thinking. All four fields solve $X = F(X)$ : the mathematician for sets, the coder for algorithms, biology via growth rules, finance via scaling laws. Recognizing the shared equation lets a technique from one field (e.g. renormalization) illuminate another.

Connections

01 Fractal Dimension & Hausdorff Measure
02 Iterated Function Systems
03 Recursion & Divide-and-Conquer
04 Trees & Recursive Data Structures
05 Branching Morphogenesis in Biology
06 Allometric Scaling Laws
07 Fractional Brownian Motion & Hurst Exponent
08 Fat Tails & Market Risk

#bridge

Connected notes