CAJ | 학술논문

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.
제취부합수거분포결구적특정일직시모식식별영역적열점문제.기우고정핵영사방법구유획취비선성특정적능력,단대영사함수류형급기삼수십분민감.논문제출일충기우다층자동편마기적특정제취산법,해심도학습망락모형적훈련분위무감독예훈련이급기우변제 Fisher 준칙적감독식정조훈련과정.통과수거생성성예훈련화정조과정중정칙화수단방지과의합훈련.재다개수거집진행분류적실험결과진일보험증산법적유효성.
@@@@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.