libsvm compute distance to hyperplane python

Some collections that relate to how to compute the distance to hyperplane in LibSVM in Python

https://github.com/bwallace/curious_snake

From Gábor Melis' () blog

Active Learning for cl-libsvm

http://mlbiomedicine.blogspot.com/2009/03/python-libsvm-or-on-hacking-libsvm.html

Some-other-source:

http://stackoverflow.com/tags/support-vector-machines/hot

http://agbs.kyb.tuebingen.mpg.de/km/bb/showthread.php?tid=358

http://agbs.kyb.tuebingen.mpg.de/km/bb/showthread.php?tid=1022

Chih-Jen Lin   Moderator Posts: 175 Joined: Sep 2004

RE: Distance to hyperplane (/computing w) from Python? (24-11-2008 05:27 PM)wallace Wrote:  One question though; I've noticed the si->obj value will at times be negative, and so therefore |w|^2 is as well. This seems rather odd to me. Am I perhaps doing something wrong? Thanks again. si->obj is  0.5 alpha^T Q alpha - sum alpha_i = -(primal obj) = -(w^Tw/2 + C \sum \xu_i) <= 0 what you should use is that alpha^TQalpha = w^Tw in Solve of svm.cpp, we have  g = (Qalpha - e)  available. By calculaing alpha^T (g+e) you get w^Tw Hope this helps.

--

♥ ¸¸.•*¨*•♫♪♪♫•*¨*•.¸¸♥