Parity-Based Amplitude Embedding for Rational Quantum Dataset Preprocessing

Quantum dataset preprocessing for amplitude embedding is understudied despite its importance for achieving good generalization in quantum machine learning models. Amplitude embedding, known for its compression capabilities, has garnered attention in recent research. Combined with the Variational Quantum Classifier, they form a potential combination for high performance in quantum machine learning, outperforming many classical models. However, the full normalization of amplitude embedding has fundamental issues, such as inconvertibility for specific values and irrational amplitude distribution due to normalization. To overcome these challenges, we propose a novel and straightforward dataset preprocessing solution using guardian parity designed for amplitude embedding. This method ensures the normalized dataset remains recoverable to its original MinMax scaled values, simultaneously enhancing dataset quality and improving accuracy when employed in quantum machine learning models. Experimental results show that the proposed parity-based amplitude embedding significantly impacts the accuracy of machine learning models compared to the original amplitude embedding method.