liveBook · Manning

Dominance

As explained in Self-Improving Agents

A relation in multi-objective optimization in which one candidate is at least as good on every objective and strictly better on at least one.

When candidates trade quality for cost, a single ranking can discard useful options. Candidate A dominates B if A is at least as good on every objective and strictly better on at least one; B can then be dropped without sacrificing a trade-off. section 4.2.3

Remove every dominated candidate and the survivors form the Pareto front: the undominated options, each best for some trade-off. Keeping the front preserves choices that a single ranked winner would lose. section 4.2.3

Quality versus cost

Among five candidates, A, B, and C remain on the front; D and E are dominated and dropped. The survivors form a staircase of trade-offs between quality and cost. section 4.2.3

Listing 4.4 The domination test and the front (02_pareto_demo.py)
# (name, quality: higher is better, cost: lower is better)
CANDIDATES = [("A", 0.90, 9), ("B", 0.80, 4), ("C", 0.60, 2),
              ("D", 0.70, 9), ("E", 0.50, 5)]
 
def dominates(x, y) -> bool:                             #A
    at_least_as_good = x[1] >= y[1] and x[2] <= y[2]
    strictly_better  = x[1] >  y[1] or  x[2] <  y[2]
    return at_least_as_good and strictly_better
 
def pareto_front(candidates):                            #B
    return [c for c in candidates
            if not any(dominates(o, c) for o in candidates
                       if o is not c)]

The code implements the two-part test—at least as good on both objectives, strictly better on one—and keeps candidates that no other candidate dominates. section 4.2.3

How Self-Improving Agents covers it

How deep it goes
Defines the pairwise test, applies it to filter a front, and illustrates the result; it does not develop a formal derivation.
Examples
Uses five quality-and-cost candidates, identifies A, B, and C as survivors, and drops dominated D and E.
What it assumes
Assumes candidates have multiple objectives with known directions, such as higher quality and lower cost.
What you can do after
Apply the domination test by hand or in code to identify survivors and design candidates that join or replace the front.
How it connects
Connects to multi-objective optimization, Pareto fronts, and evolutionary search over prompt populations.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Self-Improving Agents. The section links point to the parts of the book it draws on.