I did not solve yet but I think that the problem is related to the way I
count the length of each path.
I do it dynamically during the construction of the forest while they do it
at the end.
This seems to lead to slightly different results.
Thanks anyway :-)
On Fri, May 8, 2015 at 3:26 AM, Nico
Hi Luca
Yes try with the same parameters, and if (as I suppose) you don't have a
way to fix the same seed for the two algorithms, I suggest you to try with
large parameters to reduce the variance, like n_estimators=512 and
n_samples=200 for instance.
Nicolas
On 6 May 2015 6:57 pm, "Zay Maung Maung
Hi Luca,
Is it because of the number of trees? In R you're using 10 trees and in
python 20 trees.
Can you provide the output of your algorithm? You only provided the output
of R.
Regards,
On Thu, May 7, 2015 at 1:54 AM, Andreas Mueller wrote:
> https://github.com/scikit-learn/scikit-lear
https://github.com/scikit-learn/scikit-learn/pull/4163
On 05/06/2015 11:38 AM, Luca Puggini wrote:
Dear all,
I am wondering if there is anyone that is working on a python version
of the isolation forest algorithm.
I have written a basic draft of the algorithm but I do not understand
why I g
Dear all,
I am wondering if there is anyone that is working on a python version of
the isolation forest algorithm.
I have written a basic draft of the algorithm but I do not understand why I
get values that are different by the ones obtained with the R version.
if anyone is interested or wants t
