Expressivity limits in variational quantum portfolio construction

Abstract A recurring construction in quantum portfolio optimisation obtains weights by normalising the Pauli-\(\:Z\) expectation values \(\:⟨{Z}_{i}⟩\) of a variational circuit. We show that its standard one-layer ansatz admits the closed form \(\:⟨{Z}_{k}⟩=\prod\:_{i\le\:k}\text{c}\text{o}\text{s}{\theta\:}_{i}\), verified to \(\:4\times\:{10}^{-8}\) against simulators and complete backtests. The result generalises: in the static case, any per-qubit readout normalised into weights reparametrises the \(\:(n-1)\)-simplex and therefore admits no expressivity advantage. It admits an expressivity deficit. Because the readout is a cumulative product, \(\:|⟨{Z}_{k}⟩|\) cannot increase along the chain, and reachability reduces to whether \(\:n-1\) half-lines on a single scalar intersect: at nine assets the construction attains fewer than one portfolio in twenty thousand, and from twelve none we could sample in two million draws. Over forty training windows the circuit attains the matched classical optimum in every window where that optimum is reachable and in none where it is not (\(\:p=2.6\times\:{10}^{-7}\)), so what separates the circuit from its control is reach rather than optimisation dynamics. Depolarising noise contracts \(\:⟨Z⟩\) by \(\:(1-4p/3{)}^{k}\) and therefore degrades the portfolio toward the equal-weight portfolio, not toward randomness. The failure is invisible to the usual metrics: with the circuit 95 % of the way to equal weighting, its Sharpe ratio still exceeds the benchmark. Across five universes and 1,313 accounted configurations, none of the twenty-five quantum variants survives Holm–Bonferroni, and the largest statistic in the sweep, \(\:3.12\), falls short of the deflated threshold of \(\:3.83\). In a pre-registered sweep over eight universe sizes, the construction beats the equal-weight portfolio at all eight and the best classical competitor at none. We give the evaluation protocol that follows.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-17
DOI
https://doi.org/10.21203/rs.3.rs-11156291/v1
Primary Topic
Quantum Computing Algorithms and Architecture
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Expressivity limits in variational quantum portfolio construction

José Requena Martínez, José Manuel Zurita López
1 links
Research Square
Quantum Computing Algorithms and Architecture
preprint

Expressivity limits in variational quantum portfolio construction

José Requena Martínez, José Manuel Zurita López
preprint en

Abstract

Abstract A recurring construction in quantum portfolio optimisation obtains weights by normalising the Pauli-\(\:Z\) expectation values \(\:⟨{Z}_{i}⟩\) of a variational circuit. We show that its standard one-layer ansatz admits the closed form \(\:⟨{Z}_{k}⟩=\prod\:_{i\le\:k}\text{c}\text{o}\text{s}{\theta\:}_{i}\), verified to \(\:4\times\:{10}^{-8}\) against simulators and complete backtests. The result generalises: in the static case, any per-qubit readout normalised into weights reparametrises the \(\:(n-1)\)-simplex and therefore admits no expressivity advantage. It admits an expressivity deficit. Because the readout is a cumulative product, \(\:|⟨{Z}_{k}⟩|\) cannot increase along the chain, and reachability reduces to whether \(\:n-1\) half-lines on a single scalar intersect: at nine assets the construction attains fewer than one portfolio in twenty thousand, and from twelve none we could sample in two million draws. Over forty training windows the circuit attains the matched classical optimum in every window where that optimum is reachable and in none where it is not (\(\:p=2.6\times\:{10}^{-7}\)), so what separates the circuit from its control is reach rather than optimisation dynamics. Depolarising noise contracts \(\:⟨Z⟩\) by \(\:(1-4p/3{)}^{k}\) and therefore degrades the portfolio toward the equal-weight portfolio, not toward randomness. The failure is invisible to the usual metrics: with the circuit 95 % of the way to equal weighting, its Sharpe ratio still exceeds the benchmark. Across five universes and 1,313 accounted configurations, none of the twenty-five quantum variants survives Holm–Bonferroni, and the largest statistic in the sweep, \(\:3.12\), falls short of the deflated threshold of \(\:3.83\). In a pre-registered sweep over eight universe sizes, the construction beats the equal-weight portfolio at all eight and the best classical competitor at none. We give the evaluation protocol that follows.

Research Square
Universidad de Granada (ES)
Quantum Computing Algorithms and Architecture
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon