Research On Quantum Circuit Learning Mechanism Based On Linear Optical Quantum Computing

The emergence of quantum computing offers a new way of com-puting for dealing with problems that are intractable by classical computers and can provide exponentially accelerated computing power for efficient computation.However,the short coherence time of quantum systems,the limited interaction between quantum bits,and the crosstalk between physi-cal devices make the arrival of universal quantum computers still some time away.By optimizing the number of quantum bits,the number of quantum gates,and the depth of quantum lines,it is expected that problems of prac-tical value can be solved by using quantum computing devices in the era of noisy intermediate-scale quantum(NISQ).The hybrid quantum-classical algorithm can be used to solve problems that are difficult to be handled by classical computers using NISQ computing devices by combining classical and quantum resources to achieve the maximum efficient quantum comput-ing capability.The main research object of this paper is the quantum circuit learning mechanism based on parameterized boson sampling,which is an extension of boson sampling in the field of quantum machine learning.Since boson sampling has been little studied in solving practical problems since its intro-duction,and it still can be improved in quantum circuit structure,this paper carries out structural optimization and its corresponding application on the ground of the original boson sampling model.The parameterized boson sampling quantum circuit structure is reconstructed and designed,and then the quantum circuit learning algorithm is used to achieve quantum circuit learning efficiently by sampling the distribution of the sample set.Finally,a quantum circuit learning method based on parameterized boson sampling is proposed.The major work is shown as follows.(1)Aiming at the inflexibility of the boson sampling structure,a pa-rameterized boson sampling model based on the variational quantum cir-cuit is proposed,which improves the structure of the boson sampling and enables the updating of the qubit parameters of the parameterized boson sampling by using a suitable classical optimization algorithm and make the output result of the quantum circuit approximate the target function by a continuous iterative process? meanwhile,a fixed structure of the quantum circuit is proposed for parameterized boson sampling model,by which the complexity of the quantum circuit can be reduced and the type,number and depth of quantum gates can be optimized well.Among them,the number of parameters is only at the polynomial level of the number of photons.(2)To address the problem that boson sampling is rarely applied to other fields,a hybrid quantum-classical learning algorithm based on param-eterized boson sampling model is proposed to implement quantum circuit learning efficiently by using classical resources and optical quantum com-puting resources to expand the application field for boson sampling,and the effectiveness and reliability of parameterized boson sampling quantum circuit learning are verified by simulation experiments on the Strawberry Fields,which indicates that the algorithm has the potential to tackle practi-cal problems.