QuditML: Quantum Machine Learning using Multi-Level Systems

Quantum Machine Learning is a vibrant research field that fuses quantum computing with machine learning, taking advantage of unique quantum properties like superposition and entanglement to significantly enhance data analysis and prediction modeling. Substantial theoretical studies and initial application reports across multiple scientific domains have demonstrated the high potential of this emerging technology. To date, quantum computing primarily relies on binary quantum processing units (so called qubits), which can exist in a superposition of 0 and 1. However, many quantum computers inherently support more than these two states. A qudit uses more than two levels, enabling the processing of more information per quantum unit. Theoretical research has also demonstrated a greater variety of computational operations, potentially linked to enhanced algorithmic efficiency, improved robustness, and reduced resource requirements. Despite the immense potential, qudits remain far less researched regarding algorithmic design and utility, especially in the context of quantum machine learning. The key innovation of the project QuditML is to address this critical gap by developing novel qudit- based machine learning algorithms that leverage the computational power of multi-level quantum systems. We will explore diverse classification and regression tasks with real-world data from three scientific domains: medicine, high-energy physics, and chemistry. This will include tailored techniques for encoding multi-modal classical data (specifically tabular, signal, and imaging data) into high- dimensional, expressive qudit representations. Along with showing the first practical and algorithmic benefits, the project QuditML will also explore other previously unexamined aspects of qudit-based machine learning, such as model learning behavior, stability, and generalization performance on unseen data. This high-risk, high-reward project holds strong potential to disrupt the field of quantum computing by transforming the well-theorized advantages of qudits into the first practical machine learning algorithms. In the long term, QuditML could act as a catalyst for deeper exploration of multi-level quantum computing and contribute significantly to the advancement of quantum technologies as a whole.