Amari's 1972 associative-memory model has priority over the Hopfield network in substance — and Hopfield's 1982 paper does not cite it

amarihopfieldassociative-memorypriorityhistory-of-mlcitation

Amari's 1972 IEEE work proposed an adaptive, Hebbian-learning recurrent associative-memory network — the Lenz-Ising-derived architecture that Hopfield's 1982 PNAS paper later made famous, a decade earlier. Checked directly: Hopfield's 1982 reference list does not cite Amari 1972 (it cites two later Amari papers, 1977/1978), and Amari himself, in a 2025 interview, confirms exactly this distinction (Science Japan / JST, 2025-03-27: "My paper on associative memory was not directly cited, but a subsequent paper I wrote on self-organization was cited" — pointer and quote added 2026-09-11 audit, direct re-fetch). So Amari has substantive priority over the architecture now called the Hopfield network.

Sourcing note, added 2026-09-11 audit: the characterisation of the 1972 model above (adaptive, Hebbian, Lenz-Ising) rests on Schmidhuber's history ("In 1972, Shun-Ichi Amari made the Lenz-Ising recurrent architecture adaptive such that it could learn to associate input patterns with output patterns by changing its connection weights", Tier 2 mirror of arXiv 2212.11279) and on Wikipedia. The 1972 paper itself — "Learning Patterns and Pattern Sequences by Self-Organizing Nets of Threshold Elements", IEEE Trans. Computers C-21(11):1197–1206, DOI 10.1109/T-C.1972.223477 — has not been read directly by capture or audit (IEEE Xplore renders empty to tooling). Its abstract, via Semantic Scholar, describes nets that "remember" patterns and pattern sequences "as stable equilibrium states or state-transition sequences of the net" — consistent with, but less specific than, the Hebbian/Ising wording. Semantic Scholar lists a green open-access copy at teikyo-u.repo.nii.ac.jp/records/2067638, unread.

The claim is held with a boundary. The strongest circulating wording — "mathematically equivalent" / "exactly the same" — could not be verified from a primary reading (candidate sources: Amari's 2013 retrospective, a textbook chapter) and stays flagged [unverified-mechanism] in the capture; the load-bearing version here is the weaker, verified one: independent prior proposal, uncited by Hopfield. This is a second Amari priority thread distinct from the contested myth-amari-first-sgd-mlp (that one is single-witness and citogenesis-tainted; this one rests on Hopfield's own reference list, read directly — a stronger footing). Note the pattern recurring across the whole cluster: priority in substance, no citation in the paper that made it famous — Linnainmaa, Werbos, Parker, now Amari. See moc-backpropagation-origins, claim-reverse-mode-multiple-independent-discovery.