A Quantum Self-Attention Neural Network Model on Quantum Circuits

Area of Science:

• Quantum Computing
• Artificial Intelligence
• Natural Language Processing

Background:

• Natural Language Processing (NLP) tasks are computationally intensive.
• Quantum computing offers potential for enhanced computational power.
• Developing quantum algorithms for machine learning is an active research area.

Purpose of the Study:

• To propose a novel Quantum Self-Attention Neural Network (QSAN) model.
• To implement the QSAN model on Parameterized Quantum Circuits (PQCs).
• To evaluate the QSAN model's performance on NLP tasks.

Main Methods:

• The QSAN model architecture includes data preprocessing, quantum encoding, model design, and network optimization blocks.
• Classical text data are encoded into quantum states.
• Quantum self-attention and fully connected layers capture semantic features.
• Network parameters are optimized to enhance model performance.

Main Results:

• Simulation results show the QSAN model outperforms state-of-the-art baselines on topic classification and sentiment analysis.
• The model's effectiveness, scalability, trainability, and robustness were comprehensively evaluated.
• Current evaluations are limited to a maximum sentence length of eight words due to simulation constraints.

Conclusions:

• The proposed QSAN model presents a viable quantum computing approach for NLP.
• QSAN demonstrates potential for superior performance in advanced NLP applications.
• Further research is needed to address limitations related to sentence length and hardware constraints.