Q-PhotoMarket: A Design Space Exploration Framework for Photonic Hybrid Quantum Neural Networks in Financial Market Prediction

Alberto Marchisio; Hanzalah Mohamed Siraj; Muhammad Kashif; Nouhaila Innan; Muhammad Shafique

quant-phOct 7, 2026

Q-PhotoMarket: A Design Space Exploration Framework for Photonic Hybrid Quantum Neural Networks in Financial Market Prediction

Authors: Alberto Marchisio, Hanzalah Mohamed Siraj, Muhammad Kashif, Nouhaila Innan, Muhammad Shafique

Organizations: eBrain Lab, Division of Engineering, New York University Abu Dhabi, PO Box 129188, Abu Dhabi, UAE · Center for Quantum and Topological Systems, NYUAD Research Institute, New York University Abu Dhabi, UAE · SDU Microelectronics, Institute of Mechanical and Electrical Engineering, University of Southern Denmark, Odense, Denmark · NYUAD

Abstract

Photonic quantum computing has recently emerged as a promising platform for hybrid quantum machine learning due to its native realization of linear-optical circuits and the computational complexity of boson sampling. However, despite growing interest in quantum methods for finance, the influence of photonic circuit design choices on predictive performance remains largely unexplored. Existing studies typically evaluate a single architecture, leaving the broader photonic design space unexamined. In this work, we present Q-PhotoMarket, a systematic design space exploration (DSE) framework for photonic hybrid quantum neural networks (HQNNs) applied to financial market prediction. We explore over 5,000 valid photonic configurations spanning input photon states, circuit architectures, entangling models, and measurement strategies across their compatible computation spaces, for U.S., Indian, and cryptocurrency markets. To improve search efficiency, the exhaustive exploration is complemented with Bayesian optimization. We further incorporate threshold calibration and prediction-collapse diagnostics to enable reliable evaluation under increasingly imbalanced return thresholds. Experimental results show that systematic exploration of more than 5,000 photonic HQNN configurations reveals consistent architectural patterns across financial markets, identifies robust high-performing designs, and demonstrates competitive performance relative to classical machine learning baselines.

Figures & tables

Fig. 1: Distribution of AUC across valid photonic HQNN configurations for the S&P 500, NIFTY 50, and cryptocurrency markets.

Fig. 2: Overview of the proposed Q-PhotoMarket framework.

Fig. 3: Two-dimensional t-SNE visualization of the engineered feature spaces for the S&P 500, NIFTY 50, and cryptocurrency datasets.

Dimension	Options	Count
Input photon state	Alternating, Bunched, Pair-bunched, Thermal-like	4
Circuit architecture	Simple, Ring NN, Clements Mesh, Phase-only,	5
	Permuted Ring
Entangling model	MZI, Bell	2
Measurement strategy	Probs-Fock, Probs-Unbunched, Probs-Dual-Rail,	5
	Mode expectations, Mode correlations

TABLE I: Photonic HQNN design space explored by Q-PhotoMarket. Computation space is determined by the selected measurement strategy. Total valid configurations per (market,

\tau

) are approximately 600 after invalid-combination pruning.

Fig. 4: Prediction-collapse diagnostics. Left: calibrated decision thresholds versus positive prediction rate (PPR). Right: fraction of configurations exhibiting prediction collapse under different return thresholds.

Fig. 5: Mean AUC obtained across the explored photonic HQNN design space for different combinations of circuit architectures and measurement strategies at $\tau=1\%$ .

Fig. 6: S&P 500: 8-panel DSE diagnostics from the screening grid.

Fig. 7: NIFTY 50: 8-panel DSE diagnostics from the screening grid.

Fig. 8: Crypto: 8-panel DSE diagnostics from the screening grid.

Fig. 9: Threshold robustness of the explored photonic HQNN design space. Mean and best AUC are reported across all evaluated configurations, across

\tau\in\{0,0.5\%,1\%\}

Fig. 10: Fold-variance analysis for shortlisted configurations (Phases 1 vs 2).

R	$\tau$	Arch	Ent	Meas	State	Top10Count	MetricsHit	BestAUC	BestMCC	BestBalAcc
1	0.0100	simple	bell	probs_fock	alternating	4/8	AUC, MCC, BalAcc, GPS	0.573	0.000	0.500
2	0.0100	clements_mesh	mzi	mode_correlations	pair_bunched	4/8	MCC, BalAcc, MWA, PSharpe	0.557	0.008	0.501
3	0.0000	simple	mzi	mode_expectations	thermal_like	4/8	MCC, F1, BalAcc, AP	0.519	0.022	0.500
4	0.0000	simple	mzi	mode_expectations	bunched	3/8	MCC, F1, BalAcc	0.483	0.022	0.500
5	0.0000	ring_nn	mzi	mode_correlations	pair_bunched	3/8	MCC, F1, BalAcc	0.479	0.022	0.501
6	0.0100	clements_mesh	mzi	mode_expectations	alternating	2/8	AUC, GPS	0.576	0.000	0.500

TABLE II: Top-10 overlap table for S&P 500. Top10Count indicates in how many of 8 metrics a configuration appears in the top-10.

R	Arch	Ent	Meas	State	Top10Count	MetricsHit	BestAUC	BestMCC	BestBalAcc
1	ring_nn	bell	mode_expectations	pair_bunched	5/8	MCC, BalAcc, AP, GPS, PSharpe	0.511	0.021	0.507
2	ring_nn	mzi	mode_expectations	pair_bunched	4/8	MCC, BalAcc, GPS, PSharpe	0.516	0.048	0.522
3	ring_nn	bell	mode_expectations	bunched	4/8	MCC, BalAcc, GPS, PSharpe	0.512	0.027	0.511
4	clements_mesh	bell	mode_correlations	thermal_like	3/8	AP, GPS, PSharpe	0.520	0.003	0.500
5	clements_mesh	bell	mode_correlations	pair_bunched	3/8	AP, GPS, PSharpe	0.516	0.002	0.500
6	ring_nn	mzi	probs_fock	pair_bunched	3/8	F1, AP, GPS	0.514	0.000	0.500

TABLE III: Top-10 overlap table for NIFTY 50. Top10Count indicates in how many of 8 metrics a configuration appears in the top-10.

R	$\tau$	Arch	Ent	Meas	State	Top10Count	MetricsHit	BestAUC	BestMCC	BestBalAcc
1	0.0000	clements_mesh	mzi	mode_expectations	bunched	3/8	MCC, BalAcc, AP	0.536	0.020	0.501
2	0.0000	clements_mesh	mzi	mode_expectations	alternating	3/8	MCC, BalAcc, AP	0.536	0.031	0.503
3	0.0000	permuted_ring	bell	mode_expectations	thermal_like	3/8	MCC, F1, BalAcc	0.505	0.035	0.503
4	0.0050	clements_mesh	bell	mode_correlations	thermal_like	3/8	MCC, BalAcc, PSharpe	0.505	0.063	0.514
5	0.0050	clements_mesh	mzi	mode_correlations	thermal_like	3/8	MCC, BalAcc, PSharpe	0.503	0.042	0.509
6	0.0000	ring_nn	bell	mode_correlations	pair_bunched	3/8	MCC, F1, BalAcc	0.500	0.031	0.503

TABLE IV: Top-10 overlap table for Crypto. Top10Count indicates in how many of 8 metrics a configuration appears in the top-10.

Fig. 11: Accuracy comparison against majority, logistic, and MLP baselines.

Explore similar work

Q-PhotoNAS: Hybrid Quantum Neural Architecture Search Framework on Photonic Devices
May 21, 2026Farah Elnakhal, Alberto Marchisio, Nouhaila Innan +2Neural Architecture Search Nanophotonic
Science Division, New York University Abu Dhabi, UAE · eBrain Lab, Division of Engineering, New York University Abu Dhabi, PO Box 129188, Abu Dhabi, UAE · Center for Quantum and Topological Systems, NYUAD Research Institute, New York University Abu Dhabi, UAE +1
PN-QNN: Harnessing Physical Noise as a Native Regularizer in Photonic Hybrid Quantum Neural Networks
Jul 22, 2026Farah Elnakhal, Alberto Marchisio, Nouhaila Innan +2Quantum Neural Networks Single-Photon Avalanche Diode
Science Division, New York University Abu Dhabi, UAE · eBrain Lab, Division of Engineering, New York University Abu Dhabi, PO Box 129188, Abu Dhabi, UAE · Center for Quantum and Topological Systems, NYUAD Research Institute, New York University Abu Dhabi, UAE +1
Hybrid Quantum-Classical Neural Architecture Search
May 18, 2026Alberto Marchisio, Muhammad Kashif, Nouhaila Innan +1Quantum Neural Networks Neural Architecture Search
eBRAIN Lab, Division of Engineering, New York University Abu Dhabi (NYUAD), Abu Dhabi, United Arab Emirates · Center for Quantum and Topological Systems (CQTS), NYUAD Research Institute, NYUAD, Abu Dhabi, United Arab Emirates

May 21, 2026, quant-phJ/K move · Enter open · S save

Q-PhotoNAS: Hybrid Quantum Neural Architecture Search Framework on Photonic Devices

Farah Elnakhal, Alberto Marchisio, Nouhaila Innan, Gabriel Falcao+1

Science Division, New York University Abu Dhabi, UAE · eBrain Lab, Division of Engineering, New York University Abu Dhabi, PO Box 129188, Abu Dhabi, UAE · Center for Quantum and Topological Systems, NYUAD Research Institute, New York University Abu Dhabi, UAE+1

Neural Architecture Search Nanophotonic

Open paper AskPDF arXiv

Photonic quantum computing is a promising platform for scalable quantum machine learning, but designing effective hybrid architectures remains challenging under hardware and optimization constraints. Existing approaches rely on manually tuned architectures that fail to account for the collaboration between classical preprocessing, phase encoding, and photonic circuit structure, limiting both accuracy and hardware compatibility. In this paper, we propose a neural architecture search framework for hybrid photonic quantum-classical models that combines genetic algorithm-based search with learnable quantum phase encoding to systematically explore the joint design space of classical and quantum components. Our framework encodes 19 hyperparameters across six gene groups and evolves a population of hybrid architectures using group-based crossover, per-gene mutation, and elitism, evaluating each candidate on a short training budget before full retraining of the best found design. We evaluate our framework on two image classification benchmarks, Digits and MNIST, achieving final validation accuracies of 99.44% and 98.78%, respectively, with first-principles execution time estimates on the Quandela Ascella photonic QPU projecting single-image inference at 67 ms (Digits) and 149 ms (MNIST). Our quantum contribution analysis further shows that the photonic layer extracts non-redundant features orthogonal to the classical pathway, providing a measurable accuracy advantage over classical-only baselines. Our results demonstrate that automated architecture search is both practical and impactful for hybrid photonic systems, opening the way for systematic design space exploration of quantum AI on photonic devices.

Abstract

Explore similar work

Q-PhotoNAS: Hybrid Quantum Neural Architecture Search Framework on Photonic Devices

PN-QNN: Harnessing Physical Noise as a Native Regularizer in Photonic Hybrid Quantum Neural Networks

Hybrid Quantum-Classical Neural Architecture Search