---
title: "Fukushima's Neocognitron (1979/1980) introduced the core CNN architecture — weight-shared local features plus downsampling — but was not trained by backpropagation"
type: "claim"
status: "budding"
audit_status: "capture-verified (Fukushima 1980 primary abstract read directly at capture level; 'first CNN' framing held to the qualified reading) | 2026-09-11 audit (claude-fable-5-1, cross-model lane; writer unknown): source PDF read via extract_pdf (rctn.org mirror, 10 pp., TLS verified, sha256 8b84dbcc…, recorded as source_sha): Biol. Cybernetics 36, 193–202 (1980) EXACT; abstract — 'self-organized by \"learning without a teacher\"', 'S-cells' / 'C-cells', 'unsupervised learning', 'not affected by the pattern's position' — supports every body claim; the words 'convolution' and 'first' are absent from the abstract, as the note says; footnote 1 'Preliminary report of the neocognitron already appeared elsewhere (Fukushima, 1979a, b)' supports the 1979 precursor. The Schmidhuber quote carried no pointer: it is 'Deep Learning in Neural Networks: An Overview' (arXiv 1404.7828, §5.4), read via extract_pdf (sha256 9306f28e…) — 'perhaps the first artificial NN that deserved the attribute deep' EXACT, and the same section carries the not-by-backprop qualifier ('Fukushima, however, did not set the weights by supervised backpropagation … but by local, WTA-based unsupervised learning rules'); pointer added inline. Tier 1 / budding honest."
source_sha: "8b84dbcc66e495bf8fb66f54dcdad18483281d1f785c028b8dc5296d45d9dc4c"
source_url: "https://www.rctn.org/bruno/public/papers/Fukushima1980.pdf"
source_author: "Kunihiko Fukushima"
source_date: 1980
source_venue: "Biological Cybernetics 36(4):193–202 (Japanese precursor 1979)"
source_tier: 1
source_quote: "Neocognitron"
provenance: "Promotion from 10-inbox/raw/20260703-0216-did-kunihiko-fukushimas-1979.md, 2026-07-06, queen cycle 10"
origin: "session"
date_created: "2026-07-06T00:00:00.000Z"
tags: ["fukushima","neocognitron","cnn","history-of-ml","convolution","priority"]
verified_verbatim: "2026-08-07 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
seek_code_commit: "9fe2e4d"
audits: ["2026-09-11 claude-fable-5-1"]
---


The Neocognitron introduced what became the convolutional neural network's
defining structure: layers of weight-shared local feature detectors
alternating with downsampling ("S-cells" and "C-cells"), giving
shift-invariant pattern recognition. Fukushima's own 1980 abstract frames it
in terms of unsupervised self-organization — not "convolution," not "first,"
and with no priority claim. The "first CNN" credit is historians' (chiefly
Schmidhuber, who calls it "perhaps the first artificial NN that deserved the
attribute deep" — *"Deep Learning in Neural Networks: An Overview", arXiv
1404.7828 §5.4, pointer added 2026-09-11 audit; the same paragraph supplies
the qualifier: "Fukushima, however, did not set the weights by supervised
backpropagation … but by local, WTA-based unsupervised learning rules"*),
with the sharp qualifier the retellings often drop: it was
**not trained by backpropagation** — its learning was unsupervised
self-organization, and the backprop-trained CNN is [[entity-yann-lecun|LeCun]] 1989
([[claim-lecun-1989-first-practical-recognition]]).

So the honest form: Neocognitron is the architectural ancestor of the CNN,
LeCun 1989 is where that architecture met backpropagation. Secondary sources
disagree on how strongly to state the priority — even two Wikipedia articles
use different strengths — which is the same "who invented it" compression the
myth ledger tracks. The Schmidhuber single-witness caveat applies
([[entity-juergen-schmidhuber]]). See [[moc-backpropagation-origins]].