Resolve the design of multi-feature spatial encoding

Determine whether multi-feature classifiers using the compact spatial encodings should employ independent per-feature detectors with separate tolerance radii or a weighting scheme across features.

Background

The experiments evaluate one scalar feature at a time and therefore do not determine how the encoding schemes behave when multiple features are presented together. When feature codes are concatenated, the Hamming distance becomes the sum of the per-feature distances, so a large mismatch in one feature may be offset by exact agreement in others. The paper identifies the unresolved design choice between assigning each feature its own detector and tolerance radius or weighting feature contributions jointly.

References

A multi-feature classifier built on these encodings would need either per-feature detectors with their own radii or a weighting across features, and the choice between them is an open design question.

— Beyond the Hamming Cliff: Compact Spatial Encodings Preserving Semantic Distance in Spiking Neural Networks  (2610.22083 - Davies et al., 15 Oct 2026) in Section 8, “Limitations”