← Symmetry Survey
Symmetry Survey / Deepening / 3 of 6 / which quotient directions could SGD actually use?
Deepening 3 of 6

Transport Geometry: Surviving Fibers, Slack, and Stepwise Error

A candidate symmetry direction becomes physically interesting only if it survives three filters: better observables, realistic slack budget, and stepwise transport through the dynamics.

This is the page that turns the symmetry survey into an optimization picture. The goal is not “any invisible move,” but “a move that stays nearly invisible under the resolutions and budgets that matter.”

Theorem Reference

Lean anchors. boundaryWritableRegion_antitone, ApproximateTransportCriterion, approximateTransportCriterion_of_exact_depthGauge, refined_budgeted_step_transport_criterion, approximate_interlingua_step_transport

Math statement.

\[ \mathcal O_{\mathrm{coarse}} \subseteq \mathcal O_{\mathrm{fine}} ,\qquad D \text{ preserves }\mathcal O_{\mathrm{fine}} \text{ within } \varepsilon, \qquad c(D)\le b \] \[ \Longrightarrow D \text{ preserves }\mathcal O_{\mathrm{coarse}} \text{ within } \varepsilon, \qquad d_{t+1} \le d_t + \varepsilon_A + \varepsilon_B. \]

In English. A transport move counts as a serious candidate when it survives a finer observable test, fits inside a chosen scalar rewrite budget, and still sits inside a controlled one-step latent transport bound.

Physical intuition. Real low-loss fibers are the ones that survive refinement, remain affordable under realistic rewrite budget, and do not blow up under the actual stepwise dynamics. The new theorem package makes that filter explicit without relying on the broken legacy budget file.

physics: stable gauge fiber optimization: viable low-loss path engineering: safe scalar-budget reparameterization

Three filters on a candidate transport direction

refinement filter slack-budget filter step-transport filter only directions that survive all three filters are serious transport candidates

This is the optimization reading of the survey. Geometry is not just a quotient-space slogan. It is a filter for which rewrites stay behaviorally cheap under the constraints that matter.

How to Use It

Refinement test

If a candidate fiber disappears under a richer observable set, it was probe artifact.

Budget test

If a rewrite needs more slack than the system can supply, it is not an operational symmetry direction, even if it looks invisible on paper.

Transport test

If stepwise mismatch compounds too fast, the direction is not a robust path between nearby functions.

Feynman reading: a “possible motion” only becomes a real physical degree of freedom when the apparatus cannot see it, the system can afford it, and the dynamics does not immediately tear it apart.

DIY Fiber Filter

Slide the three tests and see when a candidate direction deserves to be called a real low-loss transport fiber rather than a pretty but useless quotient direction.

refinement
78
budget
72
stability
72