基于深度学习的边际Fisher分析特征提取算法  

Marginal Fisher Feature Extraction Algorithm Based on Deep Learning

作　　者：孙志军[1,2] 薛磊[1,2] 许阳明[1,2] 

机构地区：[1]电子工程学院,合肥230037 [2]安徽省电子制约技术重点实验室,合肥230037

出　　处：《电子与信息学报》2013年第4期805-811,共7页Journal of Electronics & Information Technology

摘　　要：提取符合数据分布结构的特征一直是模式识别领域的热点问题。基于固定核映射方法具有获取非线性特征的能力,但对映射函数类型及其参数十分敏感。论文提出一种基于多层自动编码器的特征提取算法,该深度学习网络模型的训练分为无监督预训练以及基于边际Fisher准则的监督式精雕训练过程。通过数据生成性预训练和精雕过程中正则化手段防止过拟合训练。在多个数据集进行分类的实验结果进一步验证算法的有效性。It is always important issue to extract features that are most effective for preserving the distribution architecture in pattern recognition community. Kernel based methods are assumed to extract nonlinear features. However, it is very sensitive to the selection of its mapping function and parameters. This paper proposes a feature extraction algorithm based on multi-layer auto-encoder, which consists of two phases of unsupervised pretraining and supervised fine-tuning based on marginal Fisher rule. Generative pretraining and regularization methods within fine-tuning phase are adopted to avoid overfitting of model＇s training. The validity of algorithm is proved within the result of classification experiments in several datasets.

关 键 词：模式识别 特征提取 深度学习 自动编码器 边际Fisher分析 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]