Hi,
You cannot use complex numbers, they should be real numbers. Each data
point should be in |R^d (where d is the dimensionality).
2014-09-17 19:45 GMT+02:00 Neal Becker :
> I just tried k-nearest neighbors where the data are complex. It doesn't
> seem to
> work correctly.
>
> I tried
>
> impo
I just tried k-nearest neighbors where the data are complex. It doesn't seem
to
work correctly.
I tried
import numpy as np
from const64apsk import gen_constellation_64apsk
const = gen_constellation_64apsk ('3/4')
X = [[e] for e in const]
y = np.arange(64)
from sklearn.neighbors import KNeigh
Thanks Arnaoud.
Josh
On Fri, Sep 12, 2014 at 2:03 PM, Arnaud Joly wrote:
> Here the link to the issue
> https://github.com/scikit-learn/scikit-learn/issues/3455
>
> Arnaud
>
> On 12 Sep 2014, at 20:01, Arnaud Joly wrote:
>
> If you want to work on custom oob scoring, there is an issue opened
>
Andy,
Indeed, this will mostly depend on the number of public utils we have.
However, using submodules can help structure our public utils.
M.
On Wed, Sep 17, 2014 at 6:32 PM, Andy wrote:
> On 09/15/2014 03:40 PM, Mathieu Blondel wrote:
>
>> lightning is using the following utils:
>>
>> - chec
Hi Maksym.
If you only want the loss to be reweighted according to class, you can
simply use sample_weights to give more emphasis to the samples of this
class.
If you want some other loss function, you might need to specify your own
splitting criterion.
Cheers,
Andy
On 09/16/2014 08:39 AM,
On 09/15/2014 03:40 PM, Mathieu Blondel wrote:
> lightning is using the following utils:
>
> - check_random_state
> - safe_sparse_dot
> - shuffle
> - safe_mask
> - sklearn.utils.testing.*
>
> The latter is not big deal but I like importing assertions from the
> same place.
>
> On a second thought,
I was wondering what the rationale is for making the default token pattern
for the CountVectorizer require *2* or more alphanumeric characters to form
a token. This was not intuitive default behavior for me, so I ended up with
a bug where some strings in my vocabulary like "Hepatitis A" were not
co
