en.Wedoany.com Reported - WiMi Hologram Cloud Inc. (NASDAQ: WiMi) is researching the use of neural networks to optimize parameter configurations in Twin-Field Quantum Key Distribution (TF-QKD) systems through machine learning. The company stated that this approach aims to leverage the powerful fitting and generalization capabilities of neural networks to directly predict the optimal parameters of the system, thereby significantly reducing computation time and resource consumption.
In the study, WiMi trained and evaluated three different types of neural network models: Back Propagation Neural Network (BPNN), Radial Basis Function Neural Network (RBFNN), and Generalized Regression Neural Network (GRNN). BPNN is based on the error backpropagation algorithm, continuously adjusting weights and biases to minimize prediction errors; RBFNN uses radial basis functions as activation functions in the hidden layer, suitable for handling high-dimensional data and nonlinear problems requiring high precision; GRNN is based on probability density estimation, employing kernel function methods for nonlinear regression, performing well with small sample data and uncertainty issues.
Test results showed that all three models could accurately predict the optimal parameters of the TF-QKD system to a certain extent. Among them, RBFNN and GRNN performed better in high-dimensional parameter spaces, achieving higher prediction accuracy. Compared with the LSA method, the neural network-based prediction approach can reduce computation time by several orders of magnitude. BPNN, due to its relatively simple structure, had the fastest computation speed; the computational cost of RBFNN and GRNN was slightly higher but still within an acceptable range, and their higher prediction accuracy often brought greater practical application value.
To address the differentiated requirements for real-time performance and accuracy in different TF-QKD systems, WiMi also compared prediction accuracy and time consumption. The results indicated that in scenarios requiring rapid response with lower accuracy demands, BPNN is a more suitable choice; while in applications prioritizing high accuracy and tolerating certain computation time, RBFNN or GRNN are more appropriate.
The main technical advantages of this method lie in significantly reducing the computational complexity of parameter optimization, accelerating the key generation rate, and enhancing the system's real-time response capability. Neural networks can automatically learn and adapt to changes in the quantum communication environment, enabling dynamic adjustment of system parameters. As quantum communication technology develops, this model can be further upgraded to address more complex quantum key distribution protocols and higher security requirements.
WiMi stated that it will continue to deepen research on neural networks for TF-QKD parameter optimization, exploring more advanced architectures and training strategies, such as deep learning and reinforcement learning, to build more efficient and intelligent quantum key distribution systems. At the same time, the company will strengthen integration with quantum communication hardware platforms to promote the practical application and commercialization of this technology.
