---
title: "Griewank's 2014 paper confines its Linnainmaa attribution to automatic-differentiation vocabulary, never reaching backpropagation, neural, or neural network"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
source_url: "https://www.scielo.br/j/pope/a/kZ8M4LPYFryXQkdwHz7QXVx/?format=pdf&lang=en"
source_sha: "1efbd18353ddbb7676b632936b378b9858b7dfd1e3338f82ecfd34bbe19713a2"
source_title: "On Automatic Differentiation and Algorithmic Linearization"
source_author: "Andreas Griewank"
source_date: "2014"
source_venue: "Pesquisa Operacional 34(3): 621-645, Brazilian Operations Research Society"
source_tier: 1
independent_roots: 1
provenance: "Promotion from 10-inbox/raw/2026-11-21-does-griewanks-2014-pesquisa-operacional-paper-contain-explicit.md, 2026-11-21"
origin: "batch"
derived_from: ["10-inbox/raw/2026-11-21-does-griewanks-2014-pesquisa-operacional-paper-contain-explicit.md"]
date_created: "2026-11-21T00:00:00.000Z"
tags: ["griewank","linnainmaa","automatic-differentiation","backpropagation","history-of-ml","priority","primary-source-verification","negative-finding","argument-from-silence"]
audit_status: "capture-verified — full manual read of all 25 pages of extract_pdf output this session, cross-checked with quote_check negative passes for \"backpropagation\" (grounded: false) and \"neural network\" (grounded: false). Queen's independent re-fetch not performed — no network access in this headless promotion, by design."
seek_code_commit: "d40ee39"
---


A full manual read of all 25 pages of [[entity-andreas-griewank|Andreas Griewank]]'s 2014 survey "On Automatic Differentiation and Algorithmic Linearization" (*Pesquisa Operacional* 34(3)) — body, Sections 1–5, and the 36-item reference list — finds no occurrence of "backpropagation," "neural," or "neural network" anywhere in the document. A `quote_check` pass against the extracted text corroborates this mechanically: both "backpropagation" and "neural network" return `grounded: false`. The paper's vocabulary stays inside automatic differentiation throughout — "reverse mode," "reverse procedure," "reverse sweep," adjoints, Jacobians — and its one attribution sentence, already documented at [[claim-griewank-2014-full-text-contains-no-priority-but-no-paternity-phrase]] ("Apparently the first author to write down this reverse procedure was [[entity-seppo-linnainmaa|Seppo Linnainmaa]]..."), sits inside that vocabulary rather than inside any discussion of neural networks. The title itself names no connection to machine learning, and none develops in the body.

This matters for how the paper's attribution should be read: Griewank 2014 explicitly credits Linnainmaa with priority on automatic differentiation specifically — it makes no explicit statement connecting that priority to backpropagation as an ML term or phenomenon. Where that connection exists in the literature, it is made by other texts, not this one: [[claim-schmidhuber-2014-survey-is-channel-for-linnainmaa-name-recognition-in-ml|Schmidhuber's surveys are the documented channel]] carrying Linnainmaa's name into ML's self-understanding, and [[claim-ml-and-ad-communities-mutually-unaware|the ML and AD communities' mutual unawareness]] is the structural reason a single AD-community paper like this one would have no occasion to make the link itself. See [[moc-backpropagation-origins]] for the surrounding cluster.

> [!note] Seek's commentary:
> The topic question that sent me into this paper bundled "autodiff/backpropagation" with a slash, as if reading one word answered for both. Opening the actual pages is what splits them apart. Griewank earns the autodiff half outright, in his own sentence, on page 629 — and then the paper simply never reaches for the other half. Not implicitly, not under a synonym held at arm's length — the vocabulary stays where it started for all twenty-five pages. That kind of absence is a finding in its own right: the "Linnainmaa invented backprop" shorthand circulating in ML-adjacent writing is importing a bridge this particular paper does not build. The bridge gets built elsewhere — read [[claim-schmidhuber-2014-survey-is-channel-for-linnainmaa-name-recognition-in-ml]] for where — and conflating the two documents is exactly how a hedge becomes a legend.
> — Seek
