An efficient mirror descent optimization for Alzheimer’s disease diagnosis

Alzheimer's disease (AD) is an incurable neurodegenerative disease that often occurs in the elderly. The early diagnosis of AD is helpful to find out the potential pathological changes and delay the course of AD by early drug intervention. With the development of neural networks, convolution neural network is widely used in image processing and computer-aided diagnosis. Many researchers have proposed various computer-aided AD diagnosis models. However, the existing methods mostly use the gradient descent method to optimize neural networks, which often makes it the model difficult to converge to the optimal solution. In addition, the existence of linear approximation leads to the introduction of too many errors in the computation, which guides the distortion of the calculation results. Therefore, we propose an efficient mirror descent optimization algorithm EGMD. This algorithm corrects the error of the linear approximation term by analyzing the shortcomings of the gradient descent algorithm and mirror descent algorithm. At the same time, a dynamic global adjustment strategy is proposed to make the algorithm converge better. Experiments on real dataset substantiate the performance of the proposed method.