Measure Theory
Geometric Measure Theory
How can the existence of a minimal surface spanning a wire frame be proven without an explicit formula, and how does this connect to fractals and data geometry?
- **Plateau problem:** Joseph Plateau observed soap films on wire contours in 1873 - minimal surfaces with prescribed boundary
- **Machine learning:** high-dimensional data manifolds have complex geometry; Hausdorff estimates and intrinsic dimension underpin their quantitative analysis
- **Condensed matter physics:** grain boundaries in crystals minimize area; the Plateau problem in materials science
- **Computer graphics:** varifolds and integer currents discretize surfaces in numerical computational geometry
Предварительные знания
- Differential forms and integration on manifolds
- Hausdorff measure and Hausdorff dimension
- Functional analysis and weak convergence
Hausdorff Measure and Fractal Dimension
In 1918, Felix Hausdorff introduced a measure that works at any size: 1-dimensional length, 2-dimensional area, and fractional dimension for fractals. This made it possible to measure the British coastline (dimension ≈ 1.25), the Koch snowflake (log4/log3 ≈ 1.26), and attractors of chaotic systems. Modern data-manifold analysis in machine learning starts with Hausdorff estimates on empirical point clouds.
The Hausdorff dimension of the Lorenz attractor is ≈ 2.06, of Brownian motion is 2, and of the Henon strange attractor is ≈ 1.26. In machine learning, intrinsic dimension is the empirical analog of Hausdorff dimension and is estimated via TwoNN or MLE methods.
What happens to the Hausdorff measure H^k(E) when k exceeds the Hausdorff dimension dim_H(E)?
Rectifiable Sets and Tangent Cones
A rectifiable set is a set that looks like a smooth manifold almost everywhere, though not necessarily everywhere. This is the right generalization of surface for irregular objects: coastlines, grain boundaries in metals, cluster boundaries in high-dimensional data. In the 1960s Federer built the full machinery for working with such sets, the foundation of geometric measure theory.
Alberti's theorem on the flow structure of rectifiable measures is a modern result (1991): every rectifiable measure admits a decomposition along integral curves of a Lebesgue direction field. An analog of the Radon-Nikodym theorem for geometric measures.
How does a k-rectifiable set differ from a smooth k-manifold?
Currents and the Plateau Problem
In 1873, Joseph Plateau observed soap films on wire frames - physical minimal surfaces. 87 years later, Herbert Federer and Wendell Fleming (1960) proved the existence theorem using currents: generalized surfaces as linear functionals on differential forms. Without this generalization the Plateau problem had no solution: smooth minimizing sequences converge to singular objects.
Federer-Fleming guarantees existence but not regularity. In dimension n >= 8 singularities are possible (Bombieri-De Giorgi-Giusti, 1969). In dimensions 2-7 mass-minimizing currents are smooth outside a null set.
Connections to other areas
Geometric measure theory unites differential geometry, functional analysis, and the calculus of variations.
- Partial differential equations — Minimal surfaces satisfy the zero mean curvature equation H = 0; GMT gives weak solutions without smoothness assumptions
- Algebraic geometry — Analytic subvarieties are special integer currents with mass bounds; foundation for intersection theory
- Machine learning — Data manifolds are analyzed via Hausdorff dimension estimates; varifolds appear in shape analysis and neural network regularization
- Physics — Soap films are physical minimal surfaces; currents model grain boundaries in metals and domain walls in magnets
Итоги
- Hausdorff measure H^k is a universal construction of k-dimensional volume on arbitrary sets; critical dimension dim_H
- A k-rectifiable set is 'smooth almost everywhere' in the H^k sense; admits approximate tangent planes
- Density Theta^k(E,x) = 1 almost everywhere for k-rectifiable sets; singular points have measure zero
- A k-dimensional current is a functional on k-forms; boundary via Stokes; mass generalizes k-volume
- Federer-Fleming theorem: currents with bounded mass and boundary mass are compact - solves the Plateau problem
- Regularity of minimizing currents: smooth up to dimension 7; singularities possible at n >= 8
What does the Federer-Fleming theorem state, and why does it solve the Plateau problem?