Research On Quantum Convolution Calculation Based On Convolutional Neural Network

At present,convolution is widely used in image processing and many other fields.The convolution neural network under artificial intelligence is a repeated operation,which is slow and needs a lot of calculation steps.Especially with the advent of the information revolution,the data on the Internet is growing exponentially,and the convolution neural network network structure is becoming more and more complex and the number of network layers is increasing.As a result,convolution neural network needs a lot of computing resources to process information in the usual training process.The classical convolution operation can not meet the requirements of strong artificial intelligence for today's society The calculation speed and efficiency are required.However,as a new computing mode different from classical computing,quantum computing is mainly based on the quantum mechanics method to complete the operation processing on the quantum information unit.In terms of computational efficiency,quantum computing has higher advantages than classical computing due to the superposition of quantum states and the existence of parallel computing ability.In this paper,we combine the traditional convolution calculation with quantum parallel calculation,and design a new calculation model,namely quantum convolution calculation.Quantum convolution calculation model combines the advantages of quantum and convolution.It not only has the fast operation speed of quantum parallel calculation and quantum state superposition,but also has the advantages of convolution in image processing feature enhancement and image filtering.The work of this paper is as follows:1.The classical information is quantized into quantum information.Since classical information can not be used in quantum computation,we need to encode classical information into quantum information using quantum image representation(QIm R)model.2.Design one-dimensional quantum convolution calculation model.Classical convolution calculation includes multiplication,shift and summation.Through the research of quantum gate,design the corresponding algorithm principles of quantum state tensor product calculation,quantum bit probability amplitude replacement and addition calculation,and design one-dimensional quantum convolution circuit model diagram.3.Two dimensional quantum convolution calculation model is designed on the basis of one dimension.The difference between two-dimensional quantum convolution calculation and one-dimensional quantum convolution calculation is that due to the expansion of dimensions,it is necessary to add a probability permutation module in one-dimensional circuit design and design two-dimensional quantum convolution calculation circuit diagram.