Metric Unweighted Classical Multidimensional Scaling
Based on Forrest W. Young’s notes on Torgerson’s (1952) algorithm as presented in http://forrest.psych.unc.edu/teaching/p230/Torgerson.pdf: Step 0: Make data matrix symmetric with zeros on the diagonal Step 1: Double center the data matrix (d) to obtain B by removing row and column means and adding the grand mean of the squared data Step 2: Solve B = U * L * U.T for U and L Step 3: Calculate X = U * L**(-.5)
Args: |
|
---|---|
Kwargs: |
|
Returns: | A numpy.array with ndim columns representing the multidimensionally scaled data. |