Module 10: Algebraic Topology: Invariants as Hard Constraints
Topological invariants are not regularizers — they are hard constraints that eliminate entire connected components of the wrong hypothesis space. This module develops the fundamental group and singular homology as tools for encoding physical constraints, introduces persistent homology as a computationally tractable invariant, and applies these tools to two high-stakes domains: protein fold classification via backbone knot type, and crystallographic defect stability via the fundamental group of the order parameter space. The core lesson is that topology must come before statistics in any domain where the underlying physics has topological structure.
Learning Objectives
- Define homotopy equivalence and explain why it classifies spaces up to continuous deformation.
- Compute the fundamental group π₁ for the circle, torus, and rotation group SO(3).
- Apply the classification of topological defects: stable defects correspond to elements of π₁ of the order parameter space.
- Define singular homology H_n and compute it for simple spaces using the long exact sequence.
- Explain persistent homology, construct a barcode from a filtration, and interpret birth/death intervals.
- Apply topological invariants as hard preprocessing constraints before any statistical model is fit.
Materials
Key Concepts
- Homotopy: A continuous deformation between two maps; two spaces are homotopy equivalent if each can be continuously deformed into the other — topology classifies shapes up to such deformations.
- Fundamental Group π₁(X, x₀): The group of loops based at x₀ up to homotopy; classifies 1-dimensional holes; π₁(S¹) = ℤ, π₁(SO(3)) = ℤ₂.
- Topological Defect: A stable configuration in a physical field theory classified by a non-trivial element of π₁ of the order parameter space; cannot be removed by continuous deformation.
- Burgers Vector: The topological charge of a dislocation defect in a crystal; element of π₁(SO(3)) = ℤ₂; explains why only certain defect combinations are stable.
- Singular Homology H_n: The n-th homology group; H₀ counts connected components, H₁ counts independent loops, H₂ counts enclosed voids; computed from the singular chain complex.
- Persistent Homology: Homology computed across a filtration of scales; tracks when topological features (connected components, loops, voids) are born and when they die.
- Barcode: A visual representation of persistent homology; each bar is a topological feature, its length encodes its persistence across scales; long bars = robust features.
Central Concepts from Prerequisites
- Modules 2–6: categorical language (needed for functoriality of homology)
- Linear algebra: chain complexes, kernel and image of linear maps
- Abstract algebra: groups, group homomorphisms
- Undergraduate topology helpful but not required; key definitions provided