Extendible metric MDS in Python (akin to landmark MDS). This package implements an active metric MDS method described in "Principal Component and Correspondence Analysis in R" by Dr. Herve Abdi (2017) and is thus similar to package bios2mds written in the R language. The method allows new objects to be projected onto an existing eigenbasis defined by a set of active objects and is thus useful whenever one needs to have a stable and scalable MDS transformation.
- Python >= 3.5
- NumPy >= 1.14.0
- Pandas >= 0.22.0
pip install pymmds
There is only one core object (mmds.Space) and two methods to consider:
Space.__init__(dm: pandas.DataFrame)- takes a symmetric distance matrix of active (landmark) samples and creates an MDS spaceSpace.project(dm: pandas.DataFrame)- takes a table of distances between any number of supplementary samples and all active samples and projects the former onto the initialised MDS space.
There is also a utility function mmds.read_dm that will help you read the
DMs in case you don't want to tweak pandas.read_csv yourself.
All functions and methods are well-documented. You can either use Python's help,
iPython's ? or whatever docstring tools your IDE provides to view the docs from
within your development environment or read the
Wiki page.