Optimizing CNN Using Adaptive Moment Estimation for Image Recognition
Zeyu Liu, Xiyuan Xu, Diange Fang, Dezhi Gan · 2023
Deep learning technology has been widely used in many applications. Especially in the field of image recognition, where convolutional neural networks (CNNs) have demonstrated superior performance. However, most existing papers focus on a particular model or algorithm, and lack a complete discussion on the whole process of CNN construction, training and usage. In this paper, the authors introduce a CNN optimized by Adaptive Moment Estimation (Adam) optimizer from the theoretical level to practical applications. The authors describe its concept, algorithm, model construction, model training, simulation, testing methods, and typical cases. Adam is an extension of the stochastic gradient descent algorithm, which has recently been actively used in the field of deep learning. The dataset contains two classes, aircraft and lake, with 700 photos in each category. The authors built a convolutional neural network model and trained it by using the Adam optimizer. Experimental results demonstrate that the trained network can correctly classify the images.