Handwritten picture classification method based on quantum self-learning and self-training network

Abstract

The invention discloses a handwritten picture classification method based on a quantum self-learning and self-training network. The method comprises the following steps: constructing a neural network;generating a training set; calculating the individual length of the quantum chromosome; establishing a quantum chromosome population; carrying out quantum coding on the weight value and the bias value of the neuron; obtaining an optimal neural network by using a quantum evolutionary strategy; judging whether the current evolution falls into local optimum or not, if so, carrying out full-interference crossing, or otherwise, judging whether the evolution is terminated or not; carrying out full-interference crossing; judging whether the current evolution meets a termination condition or not, ifso, outputting a classification result, or otherwise, continuing to carry out evolution iteration; and outputting classification results. The method effectively overcomes the problems that in the prior art, conventional methods are easy to fall into local optimum and excessive hyper-parameters need to be manually set, and has the advantages of high classification precision and capability of ensuring the stability and reliability of the classification results.

Description

technical field [0001] The invention belongs to the technical field of image processing, and further relates to a handwritten picture classification method based on quantum self-learning and self-training network in the technical field of image classification. The invention can be used to realize the identification and classification of handwritten pictures of different categories from the handwritten picture data sets of different handwritings from different groups of people. Background technique [0002] For handwritten image classification tasks, it has always been a very necessary task in daily life. It is widely used in signature recognition, financial check processing, financial form processing, etc. To solve this task, it often uses techniques related to neural networks. For the neural network, the most important thing is to train the weight and bias of each neuron node to obtain the parameter value with the best classification effect under the current training to ob...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to address the deficiencies of the above-mentioned prior art, and propose a handwriting picture classification method of a quantum self-learning self-training network, which is used to solve the problem that the traditional neural network training hyperparameters need to be artificially set, the search range is small, and it is easy to fall into local problems. Optimum results in low accuracy of the handwritten picture classification model, cumbersome training and cannot guarantee the reliability and stability of the classification results

Images