ABSTRACT
This chapter discusses Multidimensional Scaling (MDS), which is a manifold learning technique. Manifold learning is an approach for non-linear dimensionality reduction. Similar to PCA, MDS is yet another classical approach to dimensionality reduction that attempts to preserve the pairwise distances between the data points. That is, it attempts to find a lower dimensional nonlinear manifold that represents the data points such that it preserves the spatial distance between the points in the higher dimensional space. In this chapter, we work through the math and the underlying theory of MDS and discuss its advantages and limitations, use cases, and some examples using datasets along with tutorials for easy understanding of how the algorithm reduces the dimensionality of the data.
