In the noisy intermediate scale quantum (NISQ) era, the restricted connectivity of qubits in quantum chip makes direct execution of dual quantum gates in quantum circuits impossible. Therefore, it is of great significance to map logical quantum circuits onto quantum chips and make double quantum gates directly executable. This paper proposes a quantum circuit mapping method based on dynamic division of circuits and recombination of gate sequences, and conducts an equivalence verification of swapping rules based on ZX-calculus. The method divides the circuit into three layers dynamically, sets a moving window behind the reference gate during the mapping process, and adopts a left-greedy movement method to reorganize the gate sequence through the exchange rules, thereby reducing the number of additional gates in the mapping process. Experimental results show that, compared with existing mapping methods, the method proposed in this work requires fewer additional gates, with an average optimization rate of 24% and a maximum optimization rate of 46%.

Aiming at the lack of adaptive target selection strategy for parameterized quantum circuits in current quantum convolutional neural network models, a quantum convolutional neural network model based on the particle swarm optimization algorithm is proposed to optimize circuits automatically. The model optimizes quantum circuits by encoding the quantum circuits as particles, then uses the particle swarm optimization algorithm to search for the circuit architectures that performs well in image classification tasks. Stimulation experiments based on Fashion MNIST and MNIST datasets show that the model has strong learning ability and good generalization performance, with accuracy up to 94.7% and 99.05%, respectively. Compared to current quantum convolutional neural network models, the average classification accuracy is improved by 4.14% and 1.43% to the maximum, respectively.