Neural Quantum Embedding via Deterministic Quantum Computation with One Qubit.

Area of Science:

• Quantum Computing
• Machine Learning
• Quantum Information Science

Background:

• Optimizing quantum data embedding is crucial for maximizing machine learning performance in quantum computing.
• Traditional embedding methods face challenges in efficiently preparing quantum states for classification tasks.

Purpose of the Study:

• To propose and validate a novel neural quantum embedding (NQE) technique using deterministic quantum computation with one qubit (DQC1).
• To enhance the classification accuracy of quantum machine learning algorithms by improving the quantum data embedding process.

Main Methods:

• Developed a neural quantum embedding (NQE) approach that trains a neural network to maximize trace distance between quantum states.
• Utilized deterministic quantum computation with one qubit (DQC1) for efficient training, leveraging its suitability for ensemble quantum systems like NMR.
• Encoded handwritten images into NMR quantum processors to validate the NQE-DQC1 protocol.

Main Results:

• Demonstrated significant improvement in data distinguishability compared to traditional embedding methods.
• Achieved 98% classification accuracy using a trained NQE and a parametrized quantum circuit, outperforming traditional embedding's 54% accuracy.
• Showcased the extendability of the NQE-DQC1 protocol, allowing NMR systems for training and other platforms like superconducting circuits for subsequent tasks.

Conclusions:

• The proposed NQE-DQC1 protocol offers an efficient and effective method for classical data embedding into quantum registers.
• Ensemble quantum systems, particularly NMR, are viable platforms for NQE training, opening new avenues in quantum machine learning.
• This work paves the way for enhanced quantum machine learning applications by optimizing the critical quantum data embedding step.