摘要
本文基于深度学习堆叠自编码网络与全连接层进行了手写数字的识别的算法研究,堆叠自编码神经网络能够挖掘数据内部信息,提取有效特征;全连接层进行分类预测。在MINIST数据集和大学生手写数字数据集中,与传统机器学习方法SVM及随机森林相比取得了更好的结果达到了0.967的准确率。本文设计的算法为堆叠自编码网络应用前景提供了研究基础。
In this paper, the recognition algorithm of handwritten numerals is studied based on deep learning stack Auto-Encoder neural network and full connection layer. The Auto-Encoder neural network can mine the internal information of data and extract effective features;all connection layers are classified for prediction. Compared with traditional machine learning methods such as SVM and Random Forest, the MINIST dataset and college students’ handwritten digit dataset have achieved better results with an accuracy of 0.967. The algorithm designed in this paper provides a research basis for the application prospect of stacked Auto-Encoder neural networks.
出处
《理论数学》
2023年第2期182-188,共7页
Pure Mathematics