Multidimensional Scaling(MDS)是一种经典的数据降维方法，同时也是数据可视化的一种手段。这个问题的最早起源，是当我们仅能获得物体之间的相似性矩阵时，如何由此来重构它们的欧几里德坐标。譬如，对一个国家的许多城市而言，假如我们并不能确定它们的经纬度信息，却知道所有城市两两之间的距离，就可以通过MDS方法将这些代表相似性的距离数据，呈现在二维坐标上。-Multidimensional Scaling (MDS) is a classical data dimensionality reduction method, and it is also a means of data visualization. The earliest origin of this problem is how to reconstruct Euclidean coordinates when we can only obtain the similarity matrix between objects. For example, for many cities in a country, if we can not determine their latitude and longitude information, but know the distance between all cities, you can MDS method to represent these similarity distance data, presented in two-dimensional Coordinate.