cnn-biologically-inspired-but-technically-imprecise

IN derived (depth 1)

Created 2026-06-21T09:59:01+00:00 · Reviewed 2026-06-21T15:37:01+00:00

CNNs are biologically inspired (receptive fields from Hubel-Wiesel 1959) but technically imprecise in two ways: the "convolution" operation is actually cross-correlation, and the features are translation-equivariant rather than translation-invariant as commonly claimed.

Justifications

SL — Biological inspiration coexists with two persistent misnomers in CNN terminology

Antecedents (all must be IN):

IN cnn-biological-origin-hubel-wiesel-1959 — CNNs trace their biological inspiration to Hubel and Wiesel's 1959/1968 discovery of receptive fields in cat and monkey visual cortex, with simple and complex cell hierarchies
IN cnn-convolution-is-cross-correlation — The convolution operation in CNNs is technically cross-correlation, not mathematical convolution
IN cnn-translation-equivariant-not-invariant — CNNs produce translation-equivariant feature maps but are typically NOT translation-invariant due to downsampling/pooling — a common misconception

Dependents

These beliefs depend on this one:

IN cnn-pragmatic-not-faithful-biological-inspiration — CNNs demonstrate that effective ML architectures don't require biological fidelity — their inductive biases (local connectivity, weight sharing) capture the right structural constraints for dramatic parameter efficiency, even though the "convolution" is technically cross-correlation and the equivariance doesn't produce true invariance.
IN ml-field-assembled-from-independent-discoveries — Machine learning as a field was assembled from independent discoveries across disconnected research communities — backpropagation was independently discovered three times across 16 years, CNNs drew imprecise biological inspiration from neuroscience, and SVMs evolved incrementally over three decades in statistical learning theory.