FWIW, argsort in pure Python is possible given some list l with:
sorted(range(len(l)), key=l.__getitem__)
Sorting by best score would be:
sorted(range(len(l)), key=lambda ind: l[ind].best_score_)
On 27 May 2014 07:39, Steven Kearnes wrote:
> Thanks for your responses. I've been using a worka
Thanks for your responses. I've been using a workaround similar to Joel's
suggestion in the meantime, and it sounds like I just have to stick with
that for now. Essentially I'm doing a hyperparameter grid search, but in a
context that doesn't support GridSearchCV, so I'm dealing with multiple
indep
It's possible to put a sequence as an object into an array, but you cannot
do it with asarray or array directly. Use, for example:
a = np.empty(1)
a[0] = estimator
# alternatively:
a[:] = [estimator]
On 26 May 2014 21:17, Gilles Louppe wrote:
> Why do you want to put a random forest in a numpy
Why do you want to put a random forest in a numpy array in the first place?
Best,
Gilles
On 26 May 2014 13:11, Lars Buitinck wrote:
> 2014-05-24 0:28 GMT+02:00 Steven Kearnes :
> > a is a list of the individual DecisionTreeClassifier objects belonging to
> > the model, instead of a list contai
2014-05-24 0:28 GMT+02:00 Steven Kearnes :
> a is a list of the individual DecisionTreeClassifier objects belonging to
> the model, instead of a list containing the model itself. The same result
> occurs if I add dtype=object to np.asarray.
>
> Why is this happening? Is there a way to prevent it?
Eq. (7) shows that their method reduces to a standard SVM objective if you
prepare input data as given in Eq. (6). Using a LinearSVC will give you w
tilde in Eq. (5) and you can make predictions with Eq. (1). I said
something wrong in my previous email, the shape of X should be [n_samples,
n_featur
I did send to the author, but he does not reply.
Yes, I've implemented the normalization and k-means using scikit.
Can you please explain a bit about your proposed solution? It is completely
different from what I was thinking it should be.
The paper suggests to subtract the global classifier fro
Hi,
Have you tried to ask the authors for their source code?
If your implementation is based on scikit-learn, it could be interesting to
share it as a gist.
I had a quick look at the paper and it seems to me that you don't even need
to change liblinear. You just need to prepare a sparse matrix X
