基于流形学习的非线性维数约简方法

流形学习是一种新的非线性维数约简方法,近年来正引起可视化等领域研究者的高度重视.为加深对流形学习的理解,介绍了流形学习的基本原理,总结了其研究进展和分类方法,最后阐述了几种常用的流形学习方法的基本思想、算法步骤和各自的优缺点.通过在人工数据集Swiss-Roll上进行实验,将各类方法在近邻值选取和噪声影响等方面进行了对比分析,结果表明:与传统的线性维数约简方法相比,流形学习方法能够有效地发现观测样本的低维结构.最后对流形学习未来的研究方向作出展望,以期在这一领域取得更大进展.     

Fabrication and research of bi-functional CuNi2S4 nanosheets decorated TiO2/CuNi2S4 heterojunction photoanode for photoelectrochemical water splitting

As a traditional n-type semiconductor, TiO₂ has good UV absorption ability and stable physical and chemical properties. However, its wide band gap and low oxygen evolution reaction(OER) activity limit its application in the field of photoelectrochemical(PEC) water splitting. In this work, a type-Ⅱ TiO₂/CuNi₂S₄ heterojunction photoanode is successfully constructed, which expanded the light absorption range to visible and enhanced the OER activity. First...