Binary Classification Quantum Neural Network Model Based on Optimized Grover Algorithm.

Area of Science:

• Quantum Computing
• Machine Learning
• Quantum Neural Networks

Background:

• Grover's algorithm faces limitations with unknown target solution proportions.
• Classical classifiers struggle with certain classification tasks.
• Quantum Neural Networks (QNNs) offer potential for enhanced classification.

Purpose of the Study:

• To propose a binary quantum neural network (BQNN) model addressing limitations of Grover's algorithm for unknown solution proportions.
• To enhance classification accuracy in QNNs by optimizing the Grover algorithm.
• To investigate the performance of a BQNN in binary classification tasks.

Main Methods:

• Developed a binary quantum neural network (BQNN) model.
• Optimized Grover's algorithm using partial diffusion and a trial-and-error approach to handle unknown states.
• Applied supervised learning characteristics of QNNs for binary classification.
• Tested the model's retrieval accuracy under depolarization noise.

Main Results:

• The proposed BQNN model effectively retrieves quantum states with similar features.
• Achieved a test accuracy of 97% for BQM retrieval under specific depolarization noise conditions (0.1 rate at the 20th period).
• Demonstrated retrieval accuracy improvements of approximately 4% and 10% compared to Mean Squared Error (MSE) and Binary Cross-Entropy (BCE) respectively.

Conclusions:

• The optimized Grover algorithm-based BQNN model successfully addresses the challenge of unknown target solution proportions in quantum classification.
• The proposed method shows significant improvements in retrieval accuracy, particularly in noisy quantum environments.
• This work advances the application of QNNs for robust binary classification tasks.