Quantum Inspired Ensemble Optimisation
This R&D project, developed for a European bank, explores the application of quantum-inspired optimisation techniques to systematic equity investment strategies. The research focuses on improving the robustness and predictive power of ensemble signals by optimising the selection of predictive models through a Quadratic Unconstrained Binary Optimisation (QUBO) framework. The methodology is designed to maximise signal quality while reducing redundancy across model forecasts, creating a more diversified and resilient investment signal.
The optimisation process combines a fractional-programming approach with simulated annealing to identify the most effective ensemble configuration, while remaining compatible with future hybrid and quantum-annealing infrastructures. Empirical testing on the Morningstar Eurozone Large Cap TME universe demonstrates that the QUBO-selected ensembles consistently outperform standard greedy-selection approaches in terms of robustness and internal diversification. Over a 10-year out-of-sample evaluation period, the strategy achieved statistically significant predictive results, with a mean daily Rank_IC of 0.02484.
These improvements translated into strong portfolio performance in a long-only Top 25 market-weighted strategy benchmarked against the Eurozone Large-Cap index. From 2020 onwards, the QUBO-selected strategy delivered an annualised return of 14.37% with a Sharpe ratio of 0.75, outperforming both the benchmark and the client’s previously tested Stoxx600-based strategy. The results position quantum-inspired optimisation as the strongest-performing approach among the strategies evaluated, supporting further research into broader European universes and future applications in portfolio construction and compact index replication.