Synthetic Analprint Assay

experimentNo. 0002categoryLabs & sensesrecordingpendingdataCC BY 4.0

The fly is moved through six prescribed viewpoints over an enlarged procedural radial-fold surface. Its downstream neural responses are classified by an external nearest-centroid readout to test synthetic surface discrimination.

Setup

• A procedural radial-fold height field in a neutral laboratory scene; three synthetic identities, A, B and C.
• Controls: Enroll scan, Blind test, and Run benchmark (six enrollment scans, six held-out tests, six paired disconnected-input controls).
• Each scan follows an experimenter-controlled trajectory of six fly-centered views; the displayed trace is coverage, not the fly's internal map.
• Cameras: Arena, Fly, Top, Surface.

Measurement

• Protocol retina32-lif-v1: recorded luminance columns only; surface labels, seeds, coordinates and identities never enter the neural probe.
• Each view runs 20 simulation steps on the frozen connectome; rates from 256 predefined visual-projection and descending cells form the feature vector, excluding direct retinal input cells.
• Recognition memory is a nearest-centroid classifier outside the connectome, compared with raw-luminance and retinal-spike baselines and disconnected-input controls.

Limitations

• The first pilot scored 2/6 downstream, 2/6 raw input, 1/6 retinal and 2/6 disconnected, against nominal chance of 1/3: no demonstrated recognition advantage.
• Procedural geometry only: no human scans, identity claims or clinical conclusions, and not a reproduction of the Stanford study.
• Positioning is guided, not an autonomous exploration policy; retention means reloading the external readout, not biological memory.

Sources

• Park et al. (2020), Nature Biomedical Engineering