基于度量自编码器的表示学习研究

机器学习作为目前智能研究的主流方向,受到了越来越多研究者们的关注。数据特征对机器学习算法的性能有着至关重要的影响,而表示学习则是获取特征的重要手段。本文分析已有算法优缺点,融合多种思想,提出了有监督的度量自编码器算法,并以此为基础,致力于表示学习的研究。通过在MNIST和Fashion-MNIST数据集上的实验,并与现有算法比较,本文证明了该算法的有效性。

As the mainstream of current intelligent research, machine learning has attracted more and more researchers' attention. Data features have a crucial influence on the performance of machine learning algorithms, and the representation learning is an important means to acquire features. This paper analyzes the advantages and disadvantages of existing algorithms, integrates various ideas, and proposes a supervised metric auto-encoder algorithm. Based on this, the paper is devoted to the study of representation learning. Through the experiments on the MNIST and Fashion-MNIST datasets, and compared with the existing algorithms, the paper proves the effectiveness of the algorithm.