The order of the classes of three does not matter. Example: 1,2,3 and 3,2,1 would be considered one class. I am trying to do fisher discriminant analysis on a set of data and want to begin reducing the parameters to better help with classification. As of now I cannot tell which parameters are the most pertinent, so looping every combination through and then having MATLAB take out the LDA analysis with the MisIDs on the lower end in my algorithm will help to isolate the parameters. Can someone refresh me on the math behind this/suggest a way to code this in MATLAB?

  • 2
    $\begingroup$ This is a math question and doesn't really belong here, but the answer is $\binom{70}{3} = 54740$. Look up binomial coefficients. $\endgroup$ – Yuval Filmus Jun 29 '17 at 21:17
  • $\begingroup$ There are probably better solutions to your problem. LDA (or FLD) returns a vector, and you can isolate the components of large magnitude. $\endgroup$ – Yuval Filmus Jun 29 '17 at 21:18

That probably isn't the most effective way to figure out which parameters (features) are most informative. Instead, I suggest looking at methods for feature selection or feature ranking. There are many methods out there; it's a well-studied topic.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.