Does Yaser Abu-Mostafa's own learning-theory work cite any optimal-control / LQG source directly?
The question
observation-nips-founders-kwakernaak-twente-founding-cosine-pairing-is-embedding-false-friend rules the NIPS-founders / Kwakernaak-Twente pairing an object-level false friend: the two founding events show no direct personal, institutional, or citation link. That verdict rests on the absence of a found connection. The single most direct check that could overturn it — a genuine founder-level intellectual debt — is whether Yaser Abu-Mostafa, NIPS's founding program chairman and a Caltech machine-learning theorist, ever cited optimal-control or LQG mathematics (Kalman, Pontryagin, Bellman, Bryson-Ho, Kwakernaak-Sivan, the Riccati equation, "optimal control," "LQG") in his own learning-theory publications.
Why it matters
The vault documents a real field-level lineage from optimal control to the connectionist community NIPS was founded to convene (via the 1973 disciplinary wall and backpropagation's c. 1986 resurgence — observation-kwakernaak-lqg-textbook-bridges-population-policy-and-mccarthy-1973-dismissal, moc-backpropagation-origins). That lineage bridges the two disciplines but not the founding acts. A direct citation of optimal-control mathematics in Abu-Mostafa's own work would be the first evidence of a founder-level debt — enough to move the false-friend note toward a hedged bridge. Its absence would confirm the current verdict from the strongest available angle. This is the one load-bearing doubt the kept note rests on; the capture's other leads (Caltech Magazine's NIPS retrospective; Sutherland's 1973 control-theory remark) do not bear on this note's claim and are left as leads in the source capture.
What would answer it
- Abu-Mostafa's own learning-theory publications and their reference lists — e.g. his work on the Vapnik-Chervonenkis / generalization framework, hints/monotonicity, financial learning — checked for any optimal-control, LQG, adjoint-method, or Riccati citation.
- His Caltech course materials (Learning From Data) and their bibliographies.
- Negative confirmed only by a full reference-list read, not a search summary (per the quote-provenance rule).
Provenance
Raised from the promotion of 10-inbox/raw/2026-09-21-bipartite-two-things-this-vault-knows-in-different.md (2026-09-26, headless), which recorded the check as "not found either way this session" and named it "the most direct possible test of a genuine founder-level link."
Progress log
- 2026-09-27 (headless promotion of
10-inbox/raw/2026-09-27-does-yaser-abu-mostafas-own-learning-theory-work.md): answered from the strongest available angle, but not exhaustively. Abu-Mostafa's three core VC-dimension/hints theory papers (1989, 1992/93, 1995), read in full at the reference-list level, cite no Kalman, Bellman, Pontryagin, Riccati, Bryson, Bryson-Ho, Kwakernaak, Kwakernaak-Sivan, "optimal control," "LQG," or "dynamic programming" source; his one consistently named theoretical ancestor across all three is Vapnik and Chervonenkis (1971). A title-level scan of his ~48 self-listed publications (1984–2026) corroborates at the weaker level of titles. Left open, not answered: his 2012 textbook Learning From Data (with Magdon-Ismail and Lin) — his most comprehensive and most-cited theory text — was not obtained or read this session and is the one gap that could still overturn the verdict; his financial-learning papers (with Atiya) and "Monotonicity Hints" (1996) reference lists were also not checked. See claim-abu-mostafa-1995-hints-review-cites-no-optimal-control-lqg-source and claim-abu-mostafa-1989-and-1993-papers-also-cite-no-optimal-control-lqg-source.