Neural Embedding Geometry

Cognitive states embedded as points in a manifold.

φ: brain → ℝⁿ

Sigil ν Operator K CEFR C1 Week 8

C · CompressionK · ThresholdF · FoldU · UnfoldingG · Generation

Version — Book 3 v2 (web, dated 2026-09-25). This page is the live web edition, part of Book 3 version 2, to be deposited on Zenodo at the March 2027 equinox. Added since version 1: Lean status notes. Version 1, as taught and deposited: 10.5281/zenodo.22947908.

OrientationThe Fourth Room

Neural embedding geometry is the room where the dm³ claim is easiest to state and hardest to test. Easiest, because the objects — population activity, coherence, phase locking — are already geometric in the working vocabulary of the field. Hardest, because the manifold is inferred rather than measured.

The orbit itself is stated compactly. Neural population activity compresses into a coherence submanifold. Curvature drives toward the synchrony threshold. At κ* the system folds into a locked theta–gamma rhythm. Unfolding selects the new oscillatory regime as the stable attractor.

μ_max = -0.55 s⁻¹ \cdot ω = 0.45 rad/s \cdot β = 2.1 \cdot κ* = 0.25-0.35 canonical invariants of the neural orbit

The OrbitCompression, Threshold, Fold, Unfolding

C · CompressPopulation activity across many units collapses onto a low-dimensional coherence submanifold. This is the step the field already performs routinely under the name of dimensionality reduction — and the step where the choice of method silently fixes the geometry.

K · ThresholdCurvature in the coherence submanifold rises toward the synchrony threshold κ* ≈ 0.30. Below it, cross-frequency coupling fluctuates without committing.

F · FoldTheta–gamma locking. Observable as a discontinuous change in cross-frequency coupling within 50–200 ms — a window short enough to distinguish a fold from a drift.

U · UnfoldThe new oscillatory regime is selected as the stable attractor. The brain has committed; the previous regime is not recoverable by reversing the input.

The 50–200 ms window is the sharpest thing in this chapter. A gradual transition over seconds would be compatible with almost any model of oscillatory dynamics. A discontinuous change inside 200 ms is compatible with a fold and awkward for most alternatives.

Falsifiability

Against intracranial recording

The coherence transition should occur at κ* ≈ 0.30, observable as a discontinuous change in cross-frequency coupling within 50–200 ms.
Seizure onset should accelerate passage through κ*. If seizure trajectories cross the threshold no faster than ordinary state changes, the curvature account adds nothing.
The open prediction carried from Chapter 21: is the coherence bridge observable in direct cortical recordings at 33 Hz?

That last one deserves an explicit caution. The number 33 recurs across this corpus — as g₃₃, as χ(H*(X⁶)) = 33 — and the corpus’s own numerology sweep (WP-29) found it marked ESTABLISHED in one file while the chapter that owns it lists it as an open conjecture. A 33 Hz prediction in cortex must be motivated from the neural geometry directly, or it is a coincidence wearing a theorem’s clothes.

Lean StatusWhat Has Been Checked

Lean status (Book 3 v2, 2026-09-25) — the neural orbit. (μ_max, ω, β) = (−0.55 s⁻¹, 0.45 rad/s, 2.1) belongs to the same dm³ family as the plasma and market rooms: the transverse rate is negative for z > 0 and tends to −0.55. (neural_inFamily, neural_transverseRate_neg, neural_transverseRate_tendsto) Kernel-checked in NeuralDynamics.lean.

Lean status (Book 3 v2, 2026-09-25) — the 50–200 ms window. The timescales carried by the stated invariants are seconds, not milliseconds: the half-life ln 2/0.55 exceeds 1 s and the fold time π/0.45 exceeds 6.9 s. So the 50–200 ms window is not derived from (μ_max, ω). It needs its own derivation — for instance, that passage through the fold is faster than relaxation onto the new cycle — before it can test the normal form. (neural_halfLife_gt_one, neural_foldTime_gt, neural_timescales_exceed_window) Kernel-checked in NeuralDynamics.lean.

BridgesWhere This Connects

Book 6 · Neural Coherence — Gamma Oscillations and ConnectomesThe Vol VI treatment of the same orbit, with connectome structure rather than normal-form parameters.
Book 6 · Hopfield Networks — The Energy Function as Lyapunov ProofWhere the attractor selection step of U is made explicit as descent on an energy function.
Week 5 · Neural OscillationsThe student-facing chapter for this orbit in the 14-week program.
WP-29 · The Numerology SweepWhy the 33 Hz prediction has to be earned rather than inherited.

← PreviousMarket · The Generative Transition Next →Neural · The Manifold and Metric

Gold Standard Science · 1 of 9 tenets shown · self-assessed against EO 14303 §3(a) — not a federal certification✗(i) reproducible — no script or Lean file
✗(ii) transparent — code or sourced references missing
✗(iii) error and uncertainty — MODEL / OPEN tags or an [HONESTY] block
✓(iv) collaborative, interdisciplinary — cross-linked to other volumes
✗(v) skeptical — states what it does not claim, or leaves items open
✗(vi) falsifiable — no check that can fail
✗(vii) unbiased peer review — not independently peer reviewed
✗(viii) negative results — no negative result reported on this page
○(ix) conflicts of interest — disclosed: author-published through G6 LLC, the author's company
Scored by tools/gold_standard.py from what is observable on this page and beside it. ✓ shown · ✗ not shown · ○ disclosure. No agency has reviewed this page. What the order says and why this series follows it.