【摘要】 深度学习用于手写数字识别时,识别精度受书写风格的影响较大。针对这一问题,利用集成学习来提高深度学习对噪声的鲁棒性,提出一种基于集成深度学习的手写数字识别模型——自适应增强多层感知器。该模型以多层感知器作为基分类器,以自适应增强算法作为集成策略。在手写数字集MNIST上进行识别实验,寻优确定了该模型的主要参数,实验结果表明自适应增强多层感知器的识别精度有一定提高。更多还原

【Abstract】 When deep learning is used for the recognition of handwritten numeral recognition, the recognition accuracy is greatly affected by the writing styles of different people. In order to solve this problem, in this paper ensemble learning is used to improve the robustness of deep learning to noise, and a handwritten numeral recognition model based on ensemble deep learning is proposed, which is named as adaptive boosting multi-layer perceptron. In this model, multi-layer perceptron is used as the base classifier, and adaptive boosting is used as the integration strategy. The main parameters of the model are determined by experiments on handwritten numeral set MNIST. The experimental results show that the recognition accuracy of adaptive boosting multi-layer perceptron is improved to some extent.更多还原