mr.piron.Back to the research desk
QUANTUM FRONTIERS

The quantum frontier starts with a classical benchmark.

A grounded look at hybrid optimization, portfolio constraints, and what quantum research can actually tell us.

TECHNICAL RESEARCHAdvanced3 primary sources297 words
MR. PIRON / RESEARCH DESK2 min read
0102123ARTICLE, IN READING ORDER →↑ NUMBERED SOURCES0102123ARTICLE, IN READING ORDER →↑ NUMBERED SOURCES0102123ARTICLE, IN READING ORDER →↑ SOURCES
FIG. 0 EVIDENCE MAP · 2 SECTIONS · 3/3 SOURCES CITED · 7 CITATIONS · 1 EQUATION · 1 CODE LISTINGComputed from this article. Each bar is a section, as long as its share of the words. A curve joins a passage to a source it cites; a circle grows with the number of passages that cite it. Diamonds mark equations and squares mark code.
Sources cited, equations and code listings per section
SectionSources citedEquationsCode listings
1. Formulate discrete selection on an equal footing with classical MIQP1, 2, 310
2. Audit feasibility ratios and post-processing attribution separately1, 201

01 /Formulate discrete selection on an equal footing with classical MIQP

Portfolio construction becomes combinatorially demanding when it combines discrete cardinality rules, minimum lot thresholds, risk budgets, and turnover penalties. Quantum research explores ways to express binary selection subproblems as Quadratic Unconstrained Binary Optimization (QUBO) tasks and sample them through hybrid quantum-classical workflows [1, 2].

The credible starting point is a matched classical benchmark. IBM and Vanguard’s published research investigates variational quantum sampling for constrained portfolio construction and compares candidate quality directly against classical mixed-integer solvers [1, 3]. IBM’s multi-objective QAOA tutorial similarly demonstrates an end-to-end circuit workflow without claiming unconditional quantum advantage [2].

When mapping a constrained NN-asset cardinality problem 1⊤x=K\mathbf 1^\top x = K into an unconstrained penalty energy EA(x)E_A(x), choose AA using a bound on how much any infeasible binary state can improve the unpenalized objective relative to a feasible reference. The spread among feasible states alone is insufficient. Verify feasibility of the penalized minimizer on small instances and audit coefficient scaling before hardware sampling.

Penalized binary portfolio energy and optimality gapEQ. 1.1
EA(x)=λK2x⊤Σx−1Kμ⊤x+A(1⊤x−K)2,Δopt(x)=E0(x)−E0(x∗)∣E0(x∗)∣E_A(x)=\frac\lambda{K^2}x^\top\Sigma x-\frac1K\mu^\top x+A\left(\mathbf 1^\top x-K\right)^2,\qquad \Delta_{\mathrm{opt}}(x)=\frac{E_0(x)-E_0(x^*)}{|E_0(x^*)|}

Every valid sample satisfying 1ᵀx = K is scored on the unpenalized objective E₀(x) against the certified classical optimum x*. Use an absolute objective gap if the reference optimum is zero.

02 /Audit feasibility ratios and post-processing attribution separately

For a custom research engagement, useful deliverables include the exact QUBO-to-Ising algebraic verification, feasibility retention diagnostics, and fair comparisons of solution quality and wall-clock runtime across exact enumeration, MIQP branch-and-bound, simulated annealing, and QAOA [1, 2].

Hardware noise, circuit transpilation depth, shot budgets, and classical post-selection all shape the observed outcome. If a greedy classical repair step fixes cardinality violations after quantum sampling, we log raw pre-repair feasibility alongside post-repair optimality gaps so classical local search is never misattributed to the quantum circuit.

  • Verify QUBO linear and quadratic coefficients against brute-force enumeration on small N before scaling.
  • Report raw in-subspace feasibility probability P(1ᵀx = K) separately from post-processed objective values.
  • Include circuit compilation, parameter optimization, and shot sampling in total wall-clock resource comparisons.
python25 LINESILLUSTRATIVE · NOT EXECUTED
# Illustrative only; exact enumeration of a six-asset cardinality problem.
from itertools import combinations
import numpy as np

mu = np.array([0.07, 0.06, 0.05, 0.045, 0.035, 0.025])
sigma = np.array([
    [0.040, 0.008, 0.004, 0.002, 0.001, 0.001],
    [0.008, 0.032, 0.005, 0.003, 0.002, 0.001],
    [0.004, 0.005, 0.025, 0.004, 0.002, 0.001],
    [0.002, 0.003, 0.004, 0.020, 0.003, 0.002],
    [0.001, 0.002, 0.002, 0.003, 0.015, 0.002],
    [0.001, 0.001, 0.001, 0.002, 0.002, 0.010],
])
n, k, lam = 6, 3, 2.5

def score(bits: np.ndarray) -> float:
    return float(lam * (bits @ sigma @ bits) / (k**2) - (mu @ bits) / k)

feasible = [np.array([1 if i in comb else 0 for i in range(n)]) for comb in combinations(range(n), k)]
exact_best = min(feasible, key=score)
worst_feasible = max(feasible, key=score)
assert len(feasible) == 20
assert all(int(bits.sum()) == k for bits in feasible)
assert score(exact_best) < score(worst_feasible)
print(f"exact_optimum={score(exact_best):.6f}")
∎

Research is complete only when the assumptions, experiment, and limitations can be reconstructed.

This perspective is research commentary. It does not constitute investment advice or evidence of live strategy performance. Experimental methods require project-specific testing. Equations and code illustrate research formulations, not measured live returns.

More perspectivesExplore the engineering atlas