---
title: "Ben Recht"
type: "entity"
entity_kind: "person"
status: "hub"
canonical_name: "Ben Recht"
aliases: []
first_seen: "2026-07-07T00:00:00.000Z"
writer_model: "claude-sonnet-5"
connects_to: ["adjoint method","backpropagation","Lagrangian duality","optimal control","costate"]
seek_code_commit: "7d42f50"
---


UC Berkeley professor whose 2016 blog post "Mates of Costate" (arg min) derives, from the control-theory side, that "the method of adjoints... is equivalent to backpropagation."

Matters to this vault as the Tier-1 source for the mathematical-equivalence (not transmission) claim joining control theory's adjoint/costate recursion to backpropagation: he attaches Lagrange multipliers to a discrete-time optimal-control problem's dynamics constraint and shows the resulting costate recursion is the same structure as backpropagation's backward pass, framed explicitly as a late personal rediscovery rather than a claim of historical descent. His post is cited across several of the vault's Werbos/Isaacs/control-theory bridge-check notes as the control-theory-side half of that equivalence.

## References
- [[claim-recht-adjoints-equivalence-not-transmission]]
- [[claim-werbos-1988-equates-dynamic-feedback-with-control-theorys-time-dependent-multipliers]]
- [[claim-werbos-1988-reference-list-names-no-isaacs-bellman-pontryagin-kelley-or-bryson]]
- [[observation-rhw-docsub-feldbaum-bridge-meta-pairing-is-false-friend-lagrange-hamilton-ancestor]]