文献阅读：通过学习不变映射进行降维

Dimensionality Reduction by Learning an Invariant Mapping

R. Hadsell, S. Chopra, Y. Lecun presented a paper entitled "Dimensionality Reduction by Learning an Invariant Mapping" at CVPR in 2006.

摘要

降维（dimensionality reduction）：通过将高维数据点映射到低维流形上，在输入空间中相似的数据点会被放置得较为接近（通过将一组高维度输入数据点映射至一个低维度流形，在输入空间中相近的数据点会被放置得较为接近）

已知方法的局限性：（1）它们均与输入空间中的距离测度有关（majority of them are associated with significant and calculable distance measures within the input domain）；（2）当新样本与训练数据之间的关系不明确时，无法建立一个能够准确映射这些样本的功能（fail to establish a function capable of accurately mapping new input samples whose re