基于TensorFlow深度学习框架的卷积神经网络研究

摘要

Convolution neural network is an artificial neural network based on multi-layer supervised learning Because of the better fault tolerance,adaptability,weight-sharing and other features,Has been widely used in image recognition,object detection and other fields.this article will be based on Google artificial intelligence system TensorFlow,and construct a CNN neural network model for handwritten numeral recognition,by changing the objective function and the activation function to improve model accuracy,finally,Tensorboard is used to visualize the model.%卷积神经网络是一种基于多层监督学习的人工神经网络由于其较好的容错性、自适应性和权值共享等特点,而被广泛应用于图像识别、物体检测等领域.就将基于Google发布的人工智能系统TensorFlow,通过构建CNN卷积神经网络模型进行手写数字识别,并对目标函数和激活函数进行优化来提高模型精准度,运用Tensorboard对模型进行可视化.