Study and Implementation of Quantum-inspired Boosting Algorithms for AI-powered Financial Asset Management
This thesis explores the application of Ensemble Learning and quantum-inspired optimisation techniques to improve the performance of machine learning models in multi-label classification tasks. Developed during an internship at Axyon AI, the project focuses on Qboost, a boosting algorithm that formulates ensemble selection as a Quadratic Unconstrained Binary Optimization (QUBO) problem. By leveraging approaches inspired by quantum annealing, the research aims to enhance the ensemble building process used within Axyon AI’s machine learning pipeline, enabling a broader and more effective exploration of candidate model configurations.
The study benchmarks the proposed Qboost-based methodology against Axyon AI’s existing ensemble learning approach and a refined classical baseline. Results demonstrate that both the optimised classical method and the Qboost-based model can improve predictive performance while also reducing overfitting. The thesis highlights the potential of integrating quantum-inspired algorithms into financial machine learning workflows and represents an initial step towards incorporating advanced quantum computational methods into Axyon AI’s technological ecosystem.