The Homological Brain: Parity Principle and Amortized Inference

Abstract

Biological intelligence emerges from substrates that are slow, noisy, and energetically constrained, yet it performs rapid and coherent inference in open-ended environments. Classical computational theories, built around vector-space transformations and instantaneous error minimization, struggle to reconcile the slow timescale of synaptic plasticity with the fast timescale of perceptual synthesis. We propose a unifying framework based on algebraic topology, the Homological Brain, in which neural computation is understood as the construction and navigation of topological structure. Central to this view is the Parity Principle, a homological partition between even-dimensional scaffolds encoding stable content ($Φ$) and odd-dimensional flows encoding dynamic context ($Ψ$). Transient contextual flows are resolved through a three-stage topological trinity transformation: Search (open-chain exploration), Closure (topological cycle formation), and Condensation (collapse of validated flows into new scaffold). This process converts high-complexity recursive search (formally modeled by Savitch's Theorem in NPSPACE) into low-complexity navigation over a learned manifold (analogous to memoized Dynamic Programming in P). In this framework, topological condensation is the mechanism that transforms a ``search problem'' into a ``navigation task'', allowing the brain to amortize past inference and achieve rapid perceptual integration. This perspective unifies the Wake-Sleep cycle, episodic-to-semantic consolidation, and dual-process theories (System 1-vs-System 2), revealing the brain as a homology engine that minimizes topological complexity to transmute high-entropy sensory flux into low-entropy, invariant cognitive structure.