Selective Qubit Utilization for Optimizing Quantum Data Compression based on Quantum State Error

Designing a quantum data compression circuit with reasonable error using autoencoders can be time-consuming. Previous studies have explored qubit compression using specially designed circuits or autoencoders, with some attempts to apply it to small 16-bit black-and-white images using four qubits. However, analysis regarding the errors using general data is not discussed enough. This paper proposes a practical technique to optimize quantum compression for general data comprising hundreds of thousands of classical data points. We introduce Quantum State Error to investigate the potential for improving accuracy through qubit selection using degree reduction in quantum compression. By doing so, imperfect quantum autoencoder circuits may exhibit enhanced performance, offering improved lossy compression capabilities. Our experiments show promising results, where certain kind of circuits responds well to degree reduction by a factor of 3/4, reducing the number of faulty qubits up to 11.44%.