I am trying to understand this paper better Greedy function approximation: A gradient boosting machine, but I start having difficulty at around Equation (6) and (7).

enter image description here

What does a gradient w.s.t. to an equation mean ($\partial F(x)$), could anyone provide a concrete example showing how it could be calculated?

  • $\begingroup$ It is to a partial derivative of a vector by another vector. It will come down to calculating a matrix where each element is a partial derivative w.r.t to each of the dimensions in vector x. $\endgroup$ Oct 2, 2018 at 2:04
  • 1
    $\begingroup$ This stuff can get pretty heavy pretty quickly. If you plan to continue your investigations or studies in this area, time spent with the introductory chapter of a good multivariate Calculus text will likely spare you grief and effort. $\endgroup$
    – whuber
    Oct 2, 2018 at 14:06
  • $\begingroup$ There must be thousands of examples of computing gradients on this site: search for gradient. Your specific question about what the gradient means is answered at stats.stackexchange.com/questions/96604. I found a few more direct answers, too: see the links at the top of this page. $\endgroup$
    – whuber
    Oct 2, 2018 at 15:12


Browse other questions tagged or ask your own question.