Quantum optimisation
A quantum algorithm tested against classical integer programming on portfolio selection.
Test instanceClassical ILPQAOA (quantum)OutcomeSample 1: 8 projects, 3 resourcesScore 331; all constraints satisfiedScore 391; budget exceeded by 5QAOA result infeasibleSample 2: 12 projects, 4 resourcesScore 333; all constraints satisfiedScore 564; all 4 constraints exceededQAOA result severely infeasibleSample 3: 200 projects, 8 resourcesScore 5,311; solved in ~1.5 sDid not execute (out-of-memory at problem construction)QAOA unable to run
Findings
- 01
The classical solver returned the proven-optimal feasible answer on all three instances; on the largest, it did so in approximately 1.5 seconds.
p. 2 - 02
The quantum solver returned answers on the first two instances, but in both cases the answers violated the resource constraints — by a small margin in the 8-project case and by more than 100 percent on every constraint in the 12-project case.
p. 2 - 03
For project-portfolio selection problems of the scale and shape examined in this pilot, classical Integer Linear Programming remains the unambiguously correct technical choice.
p. 18