DCT Lexicon · Governing Dynamics
Lagrangian Dual Coordination
Lagrangian dual coordination lets units decide separately and still reach the enterprise optimum: a central price on shared resources replaces a central plan.
DCT Governing Dynamics
In depth
Formal statement
For units sharing a resource budget, dualising the coupling constraint decomposes the problem; the master problem updates the price by a (sub)gradient step.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Dual decomposition
The dual function is concave and separable across units; its maximum equals the primal optimum under convexity and Slater's condition, and subgradient ascent converges with diminishing steps. ADMM adds a proximal term and converges under weaker conditions.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In DCT
Shadow prices are the language of coordination between business units, treasury and the transformation office: a unit that sees the right price for room above the floor, or for engineers, makes the enterprise-optimal choice without seeing the whole enterprise.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Limits
Without convexity, prices alone may not coordinate (a duality gap means some optimal allocations are unsupported by any price). Lumpy decisions need a primal recovery step.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
Go next
- Fenchel–Rockafellar Duality and Dual Bounds: Coordination prices are Fenchel–Rockafellar dual variables (Pathway 6).
- Meridian GEOP: The GEOP coordinates the units' programmes into one.
- Forward–Backward SDEs and Costates: In dynamic problems the prices evolve as costates.
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
References
- Everett (1963)
- Bertsekas (1999), Nonlinear Programming
- Boyd et al. (2011), Distributed Optimization via ADMM
DCT Governing Dynamics: the element's drawer (/governing-dynamics)
In the book: Vol. II, Ch. 3: Convex Enterprise Optimization · Vol. II, Ch. 2: The General Enterprise Optimization Problem Revisited
Related: Fenchel–Rockafellar Duality and Dual Bounds · Meridian GEOP · Forward–Backward SDEs and Costates
GD.20


