I am working on a visualization manuscript that uses dimensional reduction with tSNE of a very large data set (227,573 points x 2,571 features). The reviewers would like figures comparing the tSNE projections with other common methods of dimension reduction, aside from PCA. Unfortunately, MMA does not have implementations that I can find of specific dimension reduction methods commonly compared and cited in the field [Sammon maps, local linear embedding (LLE), ISOmap to be specific], and readily available in R, Matlab, and Python.

I have written implementations for metric and non-metric multidimensional scaling (found here with code for any fellow travelers), but am hoping somebody has developed these in MMA to save hours of coding.

Does anybody know of MMA implementations/code for these methods [Sammon maps, local linear embedding (LLE), ISOmap]?

That said, I am a bit frustrated with the long lag in MMA implementation of methods in dimension reduction, graph theory (love Szabolcs iGraph implementation), and machine learning (better, but idiosyncratic in MMA 11).

  • $\begingroup$ @AntonAntonov Perhaps only "common" in the work that I do. Please forgive my extrapolation to the larger world! I also meant "common" in the sense that R, Python, and Matlab communities have all cross-coded these methods (and many others), which have been available for several years in libraries. This becomes a problem when I work mostly in MMA (and would like to continue to do so), but need to use such methods when I publish. $\endgroup$ – GraphMan Jan 29 '18 at 15:55
  • $\begingroup$ So revised. Thanks! $\endgroup$ – GraphMan Jan 29 '18 at 18:02
  • $\begingroup$ If it is implemented in MatLab, you can use MatLink to communicate with MatLab - provided that you also get access to a MatLab license. $\endgroup$ – Henrik Schumacher Feb 8 '18 at 16:53
  • 1
    $\begingroup$ Note also that there is RLink... $\endgroup$ – Henrik Schumacher Feb 8 '18 at 16:54
  • $\begingroup$ And Pythonika... $\endgroup$ – Henrik Schumacher Feb 8 '18 at 16:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.