OQC, Citi, and the NQCC collaborated through the SparQ programme to explore a hybrid quantum-AI technique using Quantum-compressed Physics-Informed Neural Networks (QPINNs) for derivative pricing, the company says. The team demonstrated a substantial reduction in key model parameters while “preserving broadly similar behaviour” in financial modelling scenarios, actively investigating the balance between efficiency and predictive performance.
This research marks a step towards scalable quantum applications in financial services by combining quantum computing with artificial intelligence to address increasingly complex computational demands. The project evaluated whether these techniques could deliver more efficient financial modelling without compromising accuracy expected by financial institutions and regulators.
Quantum-AI Integration for Derivative Pricing
Quantum-compressed Physics-Informed Neural Networks, or QPINNs, formed the core of a collaborative effort to refine derivative pricing models, indicating a focused departure from broadly applicable artificial intelligence techniques. OQC, working with Citi and the National Quantum Computing Centre, used QPINNs within the SparQ programme to investigate efficiency gains in a notoriously complex financial calculation.
This specific methodology allowed the team to explore how quantum computing could compress the data requirements of physics-informed neural networks, a type of AI already used in financial modeling. The partnership, facilitated by the SparQ programme, underscores a sustained investment in practical quantum finance applications; it wasn’t a one-off experiment, but a structured initiative designed to evaluate real-world potential.
Maintaining predictive performance is paramount, as regulatory compliance and risk management depend on reliable model outputs. The findings suggest a pathway toward scalable quantum applications in financial services, though the technology continues to mature. Citi emphasized the importance of testing quantum approaches against real industry problems, stating they aim to “ensure that the systems we build are aligned with the needs of the financial sector.” As quantum technology progresses, these early insights will be instrumental in shaping the future of financial infrastructure, potentially enabling more accurate and responsive risk management strategies.
QPINN Architecture and Quantum-Train Compression
Quantum-Train, a method of representing neural network layers with quantum states and circuits, enabled a substantial reduction in model complexity during recent experiments with derivative pricing. The team achieved this by applying the technique to a shallow neural network architecture, decreasing the number of parameters in a key layer while maintaining broadly comparable pricing results to a classical model in numerous scenarios, according to OQC.
This approach moves beyond simply applying quantum computing to existing AI; it actively reshapes the architecture itself to utilize quantum properties. The observed trade-off between model size and predictive performance is central to the advancement of scalable AI systems for computationally intensive tasks, such as financial modelling. While some increases in modelling error were noted depending on the location of compression within the network, the errors remained within the same order of magnitude for several cases, indicating a viable path toward efficiency gains.
Benchmarking Results: Model Complexity Reduction
This work builds on physics-informed neural networks, or PINNs, which embed the differential equations governing asset price evolution directly into the neural network’s training process. By integrating financial theory into the learning process, PINNs offer a more efficient approach to solving complex mathematical models common in financial markets. Applying the Quantum-Train method to these PINNs allowed for targeted compression, demonstrating that substantial reductions in model size do not necessarily equate to significant drops in predictive accuracy, a critical consideration for financial institutions.
Scalable Quantum-Enhanced Financial Modelling Potential
This research demonstrates a shift in focus from solely improving hardware to developing application-optimised quantum computing, a strategy OQC pursues to address specific industry problems. By partnering directly with financial institutions like Citi, OQC aims to bridge the gap between quantum research and real-world implementation, acknowledging that financial services represent an early sector that could benefit from quantum capabilities, the firm reports.
Many core financial problems, including portfolio optimisation and risk modelling, are computationally intensive and mathematically complex, making them ideal candidates for quantum acceleration. The team’s findings offer valuable guidance for future hybrid quantum-classical modelling efforts, suggesting that quantum representations can enhance the efficiency of large models without significant loss of predictive power.
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
