Application Of Deep Learning In Image Classification

Deep learning is a novel technology that embodies powerful advantages in image recognition.The image classification algorithm based on deep learning is mainly based on the deep learning model,training the neural network through the back propagation algorithm,fitting the nonlinear function of the image,and combining the label of the sample to achieve the classification task of the network.This paper proposes several effective methods by studying the existing deep learning models and existing optimization algorithms:First,on the classic machine learning problem,the stochastic gradient descent algorithm is studied.On the problem of fixed learning rate in deep learning,a stochastic gradient descent algorithm based on the cyclic BB step size is proposed,and the descent and convergence of the algorithm are further proved.analysis.Numerical experiments verify that the new algorithm can speed up network training and improve network classification accuracy.Secondly,by studying the classic HS conjugate gradient algorithm,a new conjugate gradient algorithm is proposed,and the global convergence of the algorithm is analyzed under certain conditions.The proposed new algorithm is applied to different deep learning models and combined with different data set analysis.The results show that the new algorithm can speed up the training of neural networks and further improve the classification accuracy of the network.