Calibration Anchor Stimuli¶

Calibration artifact — instantiates Neighborhood-Preserving Substrate Mapping

A curated set of source elements whose correct substrate location is known in advance, presented to the map so its correspondence can be pinned, checked, and re-zeroed against ground truth.

Calibration Anchor Stimuli is a designed set of reference inputs — elements whose true position in the source space is already known — used as ground truth against which the map is calibrated. Its defining move is being the known-answer fixtures: because you know in advance where each anchor should land, presenting it reveals where the map actually puts it, exposing offset, drift, and local scale error, and giving the correspondence rule fixed pins to be fitted or re-zeroed against. It is not the algorithm that builds the map and not a self-referential fidelity metric; it is the external yardstick that lets a map be checked against something outside itself.

Example¶

To map the visual field onto a subject's visual cortex, a researcher presents anchor stimuli whose retinal position is known exactly — an expanding ring and a rotating wedge, flickering at controlled eccentricities and polar angles. Because each stimulus's position is ground truth, the cortical response to it reveals which patch of cortex represents which part of the visual field, pinning the correspondence at those points. Months later the same anchors are shown again; if the responses have shifted, the map has drifted, and by how much. The anchors never build the map on their own — they are the known-location probes that let the correspondence be fixed, checked, and re-zeroed against a reference the map cannot fake.

How it works¶

Choose anchors that span the source space and its salient axes, and record each one's known source coordinate. Present them, observe where the map places each, and read the residuals: a systematic offset means the map is miscentered, spreading residuals mean local scale distortion, and a residual that grows between sessions means drift. Use those residuals to fit or re-zero the correspondence and to report distortion at each anchor. What makes it this mechanism is that it is the reference set — the ground-truth fixtures — not the layout method that consumes them.

Tuning parameters¶

Anchor density and spread — how many anchors and how evenly they cover the space; sparse anchors leave whole regions un-pinned.
Salience-weighted placement — clustering anchors where accuracy matters most, at the cost of coverage elsewhere.
Stimulus distinctness — how separable anchors are from each other and from background; poorly separable anchors give ambiguous pins.
Re-presentation cadence — one-time calibration versus periodic re-showing to catch drift.
Known-coordinate precision — how tightly each anchor's true position is specified; loose ground truth calibrates loosely.

When it helps, and when it misleads¶

Its strength is that it imports ground truth into an otherwise self-referential map, catching absolute offset and drift that internal metrics — which only ever compare the map to itself — structurally cannot see. Its limits follow from the same fact. A map can be accurate at the anchors and wrong between them, so a sparse anchor set invites overfitting the pins while the gaps sag unnoticed. And bad anchors — a mislabeled true coordinate, a non-distinct stimulus — silently miscalibrate everything downstream, since every consumer trusts them as truth. The classic misuse is reusing stale anchors after the source space itself has changed, so the map is faithfully calibrated to a world that no longer exists. The discipline is to cover the space, weight by salience, and re-validate the anchors whenever the source moves.^[1]

How it implements the components¶

This is an input artifact — a reference set — so it fills the calibration and correspondence-pinning components and none of the testing or construction ones:

map_calibration_and_anchor_set — it is the anchor set, the ground-truth fixtures the calibration runs on.
topographic_correspondence_rule — by pinning known points, it fixes and re-zeroes the source-to-substrate correspondence at ground truth.

It does not build the correspondence in the first place — that is the Elastic-Net Embedding or Self-Organizing Map Training — and it does not align two maps to each other, which is Map Registration and Alignment.

Instantiates: Neighborhood-Preserving Substrate Mapping — it supplies the external ground truth the map is calibrated and re-checked against.
Sibling mechanisms: Map Registration and Alignment · Self-Organizing Map Training · Adaptive Remeshing and Reallocation · Adjacency-Matrix Preservation Test · Boundary and Seam Regression Test · Canary Region Probe · Coverage-Hole Heatmap · Elastic-Net Embedding · Lateral-Interaction Smoothing · Local Ablation or Lesion Probe · Magnification Function · Map Fold and Collision Scan · Multidimensional Scaling Layout · Neighborhood Trustworthiness and Continuity Metric · Perturbation-Response Map · Receptive-Field Tiling Grid · Topographic Error Measure · Versioned Coordinate Atlas

Notes¶

Calibration anchors are consumed by many other mechanisms — registration, error measures, drift monitors, the trigger for remeshing — as their shared source of truth. That makes the anchor set a quiet single point of failure: an error in the anchors' assumed coordinates propagates into every check that leans on them, and unlike a map error it will not show up as an inconsistency, because everything is agreeing with the same wrong reference.

References¶

[1] Geodetic control points and benchmarks — physically monumented locations of precisely known coordinates that surveys are tied to — are the real-world analog of an anchor set. Cited as a correctly-used example of ground-truth calibration. ↩