0
$\begingroup$

Some resources such as https://online.stat.psu.edu/stat508/book/export/html/696, give the following for the quadratic discriminant function; $$ln(\pi_k)-\frac{1}{2}(x-\mu_k)^T\Sigma_k^{-1}(x-\mu_k)-ln(|\Sigma_k|^{1/2})$$

whereas 'An Introduction to Statistical Learning' omits this last term.

Why do they omit the last term? It depends on k so surely it should be relevant to the maximisation?

$\endgroup$
1
  • $\begingroup$ The resource in arxiv.org/abs/1906.02590 includes that term. It should only vanish in the case of LDA, when all covariance matrices are equal $\endgroup$
    – PedroSebe
    Oct 6, 2020 at 10:47

1 Answer 1

0
$\begingroup$

This is listed as errata for an early edition of the book (1st edition prior to the 4th printing) http://faculty.marshall.usc.edu/gareth-james/ISL/errata.html

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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