Sorry for the cross-post. I can’t tell if this is an IPython issue, a
scikit-learn issue, or a StarCluster issue.
When I try to get back results from ExtraTreesRegressor from a
load_balanced_view on StarCluster, I get:
AttributeError: 'module' object has no attribute 'SingleBlockManager'
Not s
Update: I messed up with my training set (I included a variable I shouldn’t
have) and am now getting more reasonable results (score = .634)
My question about predicting error still stands, however. I should be able to
train a classifier on the error (now that I’ve got enough that are wrong) but
I got ExtraTreesRegressor running on IPython.parallel (Pyrallel doesn’t work
for me but the example at
http://nbviewer.ipython.org/github/ogrisel/notebooks/blob/master/Distributed%20Learning%20of%20Extra%20Trees%20with%20IPython.parallel.ipynbdid).
Now I’d like to be able to predict my error (i
Extra Trees are even more random than random forests. Have a look at
the referenced papers.
To choose one vs the other you can evaluate the generalization power
via cross-validation on your data (you might also want to grid search
the optimal parameter values for max_features and min_samples_split
2014-02-08 2:25 GMT-08:00 Arnaud Joly :
>
> I have looked a bit at your code and it’s a great start. It would be easier
> to help you if you open a pull request.
+1. Don't hesitate to open an early PR with the "[WIP]" marker as a
title prefix to emphasize that you don't consider it finished work y
This looks perfect. I’m pretty knew to ensemble methods, so please forgive this
ignorant question: what’s the difference between ExtraTrees and RandomForests?
From http://scikit-learn.org/stable/modules/ensemble.html it looks like
ExtraTrees is an extension of RandomForests. Examples of when one
2014-02-08 18:44 GMT+01:00 Lorenzo Isella :
> This is the range of my data.
>
> train.max() is,
> 2.33326321223e+41
> train.min is,
> -24799.05
>
> Do you think that the max is simply too large to be handled by the random
> regressor?
Random forests cast to float32 internally, so yes, that's too l
2014-02-10 12:56 GMT+01:00 Pedro Cardoso :
> I am using the version 0.14
>
> I believe that there is a bug in defining the slices for the baches. ex: on
> a matrix qith 1078 rows, the last batch is from 1070 to 1080.
>
> Created with :
> batch_slices = list(gen_even_slices(n_batches * self.batch_si
About scikit-crab, I think the authors are re-engineering it and it is
currently not open to the community for contribution. This is what someone said
recently on the google group for crab. As far as I could tell (and I could very
well be wrong), crab recommender system does not have support for
I am using the version 0.14
I believe that there is a bug in defining the slices for the baches. ex: on
a matrix qith 1078 rows, the last batch is from 1070 to 1080.
Created with :
batch_slices = list(gen_even_slices(n_batches * self.batch_size,
n_batch
10 matches
Mail list logo
