- Stress-Based MDS
- Non-metric MDS (nMDS)
- In many real applications of MDS, dissimilarities are known only by their rank order, e.g., alphabetic order
- Weighted MDS
- Stress-Based MDS is extensible to weighted MDS in order to highlight certain aspects of dissimilarities in data
- Symbolic MDS
- MDS is extensible to dealing with symbolic data of interval of dissimilarity and distribution (histogram) of dissimilarity instead of distance between two points in metric MDS
- Interactive MDS - An extension of MDS that allows for a completely new, direct, and intuitive interaction with an MDS user on the interplay of the specific dissimilarities in data
Out of sample extension
does not provide any mapping function: z = f (x) for unseen data.
use the known X data and their embedded coordinates Z as training examples to learn a parametric mapping function e.g. Neural Networks or Support Vector Regressor