linear kernel doesnt work for gp classification

shogun-toolbox / shogun-demo

Shogun Toolbox Interactive Web Demos

7 stars 17 forks source link

Open karlnapf opened 10 years ago

karlnapf commented 10 years ago

@Saurabh7 this is for you!

Saurabh7 commented 10 years ago

@karlnapf

This is only happening with linear kernel + logit likelihood

gp.apply(test).get_values() gives [ inf -0.0105561 -0.02110647 ..., 0.09661955 0.08712393 0.07763994]

No idea what this 'inf' is doing there, that seems to be the problem.

Saurabh7 commented 10 years ago

tried to reproduce this with https://gist.github.com/Saurabh7/19ecc31591dc9a12ecf8 But that seems to work fine , sigh.

karlnapf commented 10 years ago

@yorkerlin is currently fixing bugs in our gp stuff. maybe he can have a look. @Saurabh7 could you try again to reproduce things in a script?