Finance Boost: New Portfolio Method Diversifies for Gains

Portfolio optimisation, the challenge of selecting investments to maximise returns while minimising risk, remains a crucial problem in modern finance, and researchers are continually seeking more effective solutions. José Victor S. Scursulim, Gabriel M. Langeloh, and Victor L. Beltran, all from the Instituto de Ciência e Tecnologia Itaú, alongside Samuraí Brito et al., present a new approach that directly addresses a key weakness in many existing methods: a lack of emphasis on diversification. Their work introduces a novel framework utilising a variational quantum eigensolver and specifically designed quantum states, known as Dicke states, to ensure portfolios are inherently diversified from the outset. This innovative technique significantly streamlines the optimisation process by focusing the search on only viable, diversified options, and the team demonstrates that, when paired with a particular classical optimisation algorithm, it achieves faster convergence and more accurate results, offering a promising path towards solving real-world portfolio challenges.

Application in finance represents a promising area for quantum computation, yet many proposed quantum algorithms fail to address a crucial characteristic of real-world portfolios: diversification. This work introduces a novel quantum framework for multiclass portfolio optimization that explicitly incorporates diversification by leveraging multiple parametrized Dicke states. These states are simultaneously initialized to encode the diversification constraints, functioning as an ansatz within the Variational Quantum Eigensolver. The team investigated methods to improve performance on real quantum hardware, which is often affected by errors, employing various optimization techniques and strategies to mitigate these errors. They explored techniques like Zero Noise Extrapolation and Probabilistic Error Cancellation to reduce the impact of noise on calculations. Results showed high fidelity between states generated by optimal parameters, indicating consistent performance, and the team plans to investigate the optimization landscape further, potentially using fidelity information to design a neural network for better initial parameter selection.

Quantum Portfolio Diversification via Dicke States

Researchers have developed a new approach to portfolio optimization, a critical task in finance, that explicitly addresses the need for diversification. This new method leverages the principles of quantum computing, specifically a technique called Variational Quantum Eigensolver, to efficiently explore possible investment combinations. The core innovation lies in the design of a specialized quantum circuit, built using Dicke states, which inherently enforces diversification constraints. Unlike conventional approaches, this method initializes the quantum system to consider only feasible, diversified portfolios, significantly reducing computational effort. The team demonstrated that when combined with a classical optimization algorithm, the Dicke state-based approach outperforms existing methods, converging faster, achieving a higher approximation ratio, and exhibiting a greater probability of identifying a valid portfolio. This advancement could lead to more efficient resource allocation and improved risk management strategies in the financial sector.

Diversification Improves Quantum Portfolio Optimisation

This work introduces a novel approach to multiclass portfolio optimization, employing parameterized Dicke states within a Variational Quantum Eigensolver (VQE) framework. While simulations demonstrate the viability of this approach, further research is needed to fully understand the impact of variational quantum algorithms on commercially significant problems and to test these findings on current quantum hardware. Future work will focus on exploring the Quantum Approximate Optimization Algorithm (QAOA) with Dicke states and investigating error mitigation techniques such as Zero-Noise Extrapolation and Probabilistic Error Cancellation.