MDS and LMDS implementation

An implementation of the paper Sparse multidimensional scaling using landmark points. by Vin De Silva and Joshua B. Tenenbaum.

TL;DR: MDS can be used (among other things) to project data to lower dimensions but it needs a distance matrix of NxN, where N is the number of samples. LMDS can project using a matrix of LxN, where L is the number of landmark points and L<<N.

Files

• mds.py: MDS and LMDS implementation;
• MDS_LMDS_example: notebook with usage examples.

Please contact "ddanilomotta@gmail.com" if you have any questions.