I think that the \(\lambda\) comes from the fact that he put the regularization hyperparameter in front of the regularization term.
So basically, \(\frac{\lambda}{2}\) instead of \(\frac{1}{2}\) that's why there is a lambda.
Why in this problem the update rule doesn't correspond to a SGD step ?
K_means clustering 2016
Hey,
I guess the solution is "obvious" since it is not given, but I'd be grateful if someone could share as I am quite uncertain of my answer :)
Thanks in advance
1
6
Hi, I don't get where the lambda comes from? Thanks in advance
I had a 1 in the denominator instead of lambda, i'm also unsure why it is there
2
shouldn't it be \(z_{kn}\) instead of \(z_{nk}\) ? I also have a 1 instead of lambda
1
Anyone care to share their solution for this ?
I don't get where the 1 in the denominator (or the lambda) would come from?
This is not very clean sorry
3
I think that the \(\lambda\) comes from the fact that he put the regularization hyperparameter in front of the regularization term.
So basically, \(\frac{\lambda}{2}\) instead of \(\frac{1}{2}\) that's why there is a lambda.
Why in this problem the update rule doesn't correspond to a SGD step ?
Thanks
1
Add comment