Quantum Finance Boosts Portfolios with 0.49% Error

Constructing investment portfolios that balance risk and reward while adhering to practical limitations represents a significant challenge in modern finance. Gabriele and colleagues, from their investigations into Exchange Traded Fund design, demonstrate a novel approach using quantum computing to tackle this complex optimisation problem. The team evaluates a sampling-based algorithm, combined with conventional computing techniques, to solve portfolio construction instances that quickly become intractable for even the most powerful classical computers. Their experiments, utilising circuits with over 100 qubits, achieve a remarkably low error rate of less than half a percent, suggesting that this hybrid quantum-classical workflow offers improved accuracy over purely classical methods and that more complex quantum circuits can actually enhance the solution process. This work represents a crucial step towards realising the potential of quantum computers to revolutionise financial modelling and portfolio management.

Quantum Portfolio Optimization with Constraints

Researchers are actively exploring how quantum computing can address complex financial challenges, specifically in the area of portfolio optimization. Building an investment portfolio involves maximizing returns while carefully managing risk, a task that becomes increasingly difficult as the number of potential investments grows. Traditional methods often struggle with these large-scale problems, requiring significant computational time to find even acceptable solutions. This research investigates whether quantum algorithms can offer a performance advantage over classical techniques, particularly when dealing with complex portfolios.

The team focused on utilizing Conditional Value at Risk (CVaR) as the primary objective for optimization. CVaR is a sophisticated risk measure that concentrates on the potential for extreme losses, making it particularly suitable for investors who prioritize minimizing downside risk. They implemented a quantum algorithm, the CVaR-based Variational Quantum Algorithm (VQA), and benchmarked its performance against established classical optimization methods. This comparison aims to determine whether quantum computing can provide a tangible benefit in solving these complex financial problems. A key aspect of the research involved investigating the scalability of these quantum algorithms, assessing how well they perform as the size of the portfolio increases. The team also addressed the impact of noise inherent in quantum computers and explored techniques to mitigate its effects on the optimization results. Creating optimal portfolios is a computationally intensive task, becoming exceedingly difficult as the number of potential investments increases. Traditional methods struggle with problems involving thousands of assets, often taking significant time to find even reasonable solutions. This work demonstrates a promising step towards overcoming these limitations by combining quantum algorithms with classical optimization techniques.

The team investigated a specific quantum approach, the Conditional Value at Risk-based Variational Quantum Algorithm (CVaR-VQA), to address the complexities of portfolio construction. This algorithm offers a unique advantage by allowing researchers to define customized cost functions, bypassing the need to convert the problem into a standard format required by many other quantum algorithms, and reducing the number of qubits needed. This flexibility allows for a more natural representation of the financial problem, potentially leading to improved solutions. A key innovation was a new method for formulating the portfolio optimization problem, specifically tailored for this quantum sampling-based approach.

Experiments involving circuits with over 100 qubits, run on IBM’s Heron processors, yielded impressive results. The combined quantum-classical workflow achieved a solution error of just 0. 49%, demonstrating a significant improvement in accuracy compared to relying solely on classical local search methods. Interestingly, the research suggests that utilizing more complex quantum circuits, those harder for classical computers to simulate, can actually lead to better convergence and more effective optimization. Furthermore, the team found that combining the quantum algorithm with a classical post-processing step, involving local search, consistently outperformed either method used in isolation. These findings pave the way for further exploration of quantum computing in finance, potentially enabling the creation of more efficient and robust investment strategies.

Quantum Portfolio Optimisation Outperforms Classical Search

This research demonstrates the feasibility of using a combination of CVaR-VQA and local search to address a realistic, simplified portfolio construction problem. Experiments involving up to 109 qubits show that this quantum-classical workflow achieves improved accuracy compared to purely classical local search methods, with a relative solution error of 0. 49% observed in the best performing circuits. Notably, the results suggest that more complex quantum circuits may converge towards better solutions, potentially indicating a regime where quantum approaches can outperform classical techniques. The authors acknowledge that scaling these quantum methods to significantly larger problem sizes, where classical solvers struggle, remains a key challenge. Future research will focus on exploring techniques to mitigate the computational demands of quantum-classical training, such as parameter transfer methods and classical-only training modes, to enable the application of these techniques to larger and more complex portfolios.