Every number on this page is computed by the Volume II Chapter 1 notebook's own code, on Volume I's 100,000 paths (seed 8); fresh paths use the lab's seed 26201. Your notes stay on this page until you leave it; copy the memo before you go.
Meridian's board wants to know what an optimizer adds to the analysis it already has, and its chief financial officer must turn Volume I's answer to the risk appetite into a package the bank and the program office can sign. You support the chief financial officer.
Use AXIOM to write Meridian's decision in canonical form, run the readiness gates, see how lumpy levers and re-weighted objectives change the answer, and document what the answer certifies.
Worked Example 1.1: Volume I's answer (a $170 million standby line and a two-month delay, $10.4 million) meets two facts: the bank writes committed lines in tranches of $50 million, and the program office plans the build in whole quarters. Breach odds are counted on Volume I's 100,000 simulated paths (seed 8).
In the formulation assistant, enter the two levers with their owners and prices, the cost objective, and the board's appetite, and read the canonical form it emits. Delete the declaration of the delay cost (its price and its owner) and watch the assistant refuse.
The canonical form (Equation 1.1): the cost with its sign reversed, over the box, subject to the state map and the one commitment. A term that cannot be evaluated from the declaration defines no optimization problem (Proposition 1.1).
Run (R1)–(R4) on the declaration. Then remove the standby line from the decision inventory and rerun (R2): which commitments does the minimal infeasible set name?
In the decision-space explorer, switch the line to $50 million tranches and the delay to whole quarters. Watch the convexity indicator, count the packages, and find the cheapest feasible one. Compare it with the continuous answer and with the two roundings.
Above the 5 percent curve every package meets the board's appetite; the dashed line joins the packages that cost the same as the best one. Dots: the $50 million tranches with whole quarters, filled where the appetite is met; the grid of $1 million and $10 million tranches is too fine to draw.
In the objective explorer, state the costs after tax and check that the answer stands. Then move the value of a quarter of delay from $8 million to $4 million and $16 million, with continuous and with lumpy levers.
The six cheapest of the 46 lumpy packages that meet the appetite, ranked by the objective in the chosen units (after tax at 25 percent: 0.75 of the cost).
The cost of the packages on the 5 percent curve, line by line, at $4, $8 and $16 million a quarter; dots mark each continuous best; the ring marks the continuous best at the slider's value.
Not about Meridian: a budget of 100 split between two channels with returns 3√x1 and 4√x2. At the optimum the marginal returns are equal, and their common value is the budget's multiplier.
In the constraint workbench, add the capability floor of 55 and the covenant with the line fully drawn, and confirm that the feasible packages can only shrink.
Each estimate on 100,000 paths has a standard error of √(p(1 − p)/N). Fresh paths: the same model drawn with the lab's seed 26201.
The refusal when a term is not declared (Proposition 1.1; gate (R1) of Theorem 1.5); feasible packages that can only shrink as commitments are added (Proposition 1.2); the continuous answer as a floor that no lumpy package beats (Theorem 1.3(v)); and the difference between a rescaling, which moves no decision, and a reweighting, which can (Theorem 1.4).
The package to sign, its odds and cost, what the lumps cost, which declaration moves the answer most, and the certificates that support each number.