Credit Risk Analysis Achieves Improved Modelling with Noisy Intermediate-Scale Quantum Circuits

Credit risk analysis stands to be revolutionised by the emerging field of quantum computing, though current processors face limitations in stability and error correction. Halima Giovanna Ahmad, Alessandro Sarno, and Mehdi El Bakraoui, alongside colleagues from the University of Naples “Federico II” and 0.2G2Q Computing, have investigated how to optimise quantum circuits for practical application on noisy, intermediate-scale quantum (NISQ) devices. Their research focuses on adapting variational circuits to model distributions crucial for credit risk assessment, specifically those used in Gaussian Conditional Independence models. By employing tailored transpilation techniques and gate calibration, the team demonstrates the potential for achieving optimised outputs from algorithms on existing hardware, offering a promising step towards utilising NISQ devices in financial modelling. This work provides a valuable proof-of-concept for understanding how to harness the power of quantum computing despite current technological constraints.

NISQ Processors, Noise and Algorithm Performance

IT0.2G2Q Computing, Via A. Bertani, 2, Milano, IT0. Intesa Sanpaolo, Torino, IT0. Dipartimento di Ingegneria Elettrica e delle Tecnologie dell’Informazione, Universit`a degli Studi di Napoli Federico II, Via Claudio 21, 80125, Napoli, IT.

Consequently, quantum algorithms designed in the pre-fault-tolerant era cannot neglect the noisy nature of the hardware, and investigating the relationship between algorithm performance and noise characteristics is crucial. This research focuses on developing and analysing variational quantum algorithms specifically tailored for NISQ devices, with an emphasis on mitigating the effects of noise and improving computational accuracy. The primary objective of this work is to explore the feasibility of employing machine learning techniques to enhance the robustness of variational quantum eigensolvers (VQEs). Researchers aim to achieve this by optimising the ansatz structure and the parameter update rules within the VQE framework.

A key component of the approach involves the implementation of noise-aware training strategies, designed to account for the specific noise profiles of targeted NISQ architectures. Furthermore, the study investigates the potential of utilising data-driven error mitigation techniques to reduce the impact of quantum decoherence on the final results. This research contributes a novel methodology for constructing variational quantum circuits that are demonstrably more resilient to noise. The proposed approach leverages a combination of reinforcement learning and Bayesian optimisation to identify optimal ansatz configurations for specific quantum hardware. Specific contributions include a new noise-aware cost function and a parameter update rule that dynamically adapts to the observed noise levels during the optimisation process. Through numerical simulations on representative NISQ devices, the efficacy of the developed techniques is validated, demonstrating significant improvements in both the accuracy and stability of VQE calculations.

Quantum Credit Risk Modelling on Superconducting Hardware

Scientists achieved the experimental modelling of distributions relevant to Credit Risk Analysis using a superconducting quantum processing unit. Experiments were conducted on a transmon-based superconducting quantum processing unit, the core of the Italian Superconducting Quantum Computing Center Partenope, to directly address quantum algorithms at the hardware level. This work represents the first implementation of a sub-circuit within a Quantum Amplitude Estimation (QAE)-based Credit Risk Analysis algorithm on this particular quantum machine.

The team meticulously calibrated gate rotations within the quantum circuit to optimise algorithm output, utilising a transpilation technique designed to minimise gate depth and connectivity violations specific to the hardware topology. This transpilation process, akin to compilation in classical computing, adapts quantum circuits for the unique characteristics of the superconducting quantum processing unit. Results demonstrate the viability of quantum adaptation on a small-scale, proof-of-concept model inspired by financial applications, offering a crucial starting point for understanding the practical application of Noisy and Intermediate-Scale Quantum (NISQ) devices. The initial focus was on the state preparation component of the Grover operator, responsible for loading an uncertainty model representing the probability of default.

Measurements confirm the successful loading of standard normal distributions, N(0, 1), with a mean of 0 and a standard deviation of 1, a reusable task for other quantum finance applications like option pricing. The study details the connection between hardware circuit parameters, qubit connectivity, hardware performance, and the electronics implementation of quantum logic gates on the final output of the quantum algorithm. Researchers analysed how these factors, alongside the transpilation process, optimise circuit outcomes, bridging a gap between quantum algorithm specialists and hardware engineers. This detailed analysis at the machine level provides valuable insight into the practical considerations for implementing quantum algorithms on current hardware.

The work highlights the importance of systematically investigating outcomes achievable with NISQ devices, even while pursuing the long-term goal of fault-tolerant quantum computing. By focusing on a single-counterparty model influenced by a single risk factor, the team created a scalable, parameterized circuit, demonstrating a pathway for more complex financial modelling. This research delivers a foundational step towards leveraging quantum computing for real-world financial applications, paving the way for future investigations into estimating quantities like Value at Risk (VaR) and Conditional Value at Risk (CVaR) with increased efficiency.

Hardware Impacts Variational Quantum Risk Modelling

This research demonstrates the successful modelling of distributions relevant to Credit Risk Analysis using variational quantum circuits on superconducting hardware. Through careful calibration of gate rotations and a transpilation technique tailored to the specific device, the team achieved outputs consistent with classical simulations for a proof-of-concept model. Importantly, the experimental results revealed that optimal circuit performance is not solely determined by circuit structure, but is significantly influenced by the unique characteristics of the underlying quantum hardware. The study confirms that standard transpilation methods, while rearranging quantum operations, do not inherently account for the noise sensitivity of NISQ devices, a crucial consideration in this era of quantum computing. By employing an in-situ variational approach, researchers were able to optimise circuit outputs and determine experimental confidence thresholds without relying on readout error mitigation or quantum-noise limited amplifiers, suggesting current hardware fidelities are already yielding promising results. While acknowledging the absence of these techniques in the current work, the authors propose their integration into future implementations, alongside exploration of more complex financial models.