Reproduce “Evaluating Angle and Amplitude Encoding Strategies for Variational Quantum Machine Learning: their impact on model's accuracy” on OpenResearch

About this paper

Antonio Tudisco, Andrea Marchesin, Maurizio Zamboni, Mariagrazia Graziano, Giovanna Turvani

This research systematically evaluates how different classical data encoding strategies, including 20 distinct Angle encoding variations and Amplitude encoding, impact the accuracy and computational cost of Variational Quantum Circuits (VQCs) for classification tasks. The study demonstrates that encoding choice is a critical hyperparameter, with optimal strategies varying significantly across real-world datasets and influencing performance by up to 41.1%.

Key insights

• Quantum encoding is confirmed as a critical hyperparameter for VQC models, with its specific choice profoundly influencing classification performance and computational efficiency.
• The optimal encoding strategy is highly dataset-dependent; no single technique universally guarantees the best classification across different datasets or evaluation metrics.
• Amplitude encoding and more complex Angle encoding variants, while potentially more expressive, consistently required significantly longer training and inference times compared to simpler Angle encoding circuits.

Summary by alphaXiv.