a sample_weight param seems reasonable to me
Alex
On Wed, Apr 11, 2012 at 5:10 PM, Olivier Grisel
wrote:
> Le 11 avril 2012 16:59, Michael Selik a écrit :
>> Certainly. It looks like a good approach would be to break out line 121 in
>> mean_shift_.py:
>>> my_mean = np.mean(points_within, axis=
of potential interest to the hmm module.
cheers,
satra
-- Forwarded message --
From: George H. Chen
Date: Mon, Apr 9, 2012 at 9:10 PM
Subject: [ML tea] Matt Johnson's writeup for his talk today on spectral
learning of HMM's
Hello all.
To those who missed it and for those who
Le 11 avril 2012 16:59, Michael Selik a écrit :
> Certainly. It looks like a good approach would be to break out line 121 in
> mean_shift_.py:
>> my_mean = np.mean(points_within, axis=0)
>
> And provide a function instead that allows several methods of mean
> calculation -- flat kernel (current
Certainly. It looks like a good approach would be to break out line 121 in
mean_shift_.py:
> my_mean = np.mean(points_within, axis=0)
And provide a function instead that allows several methods of mean calculation
-- flat kernel (current method), gaussian kernel, and/or accuracy-weighted
kernel.
On Wed, Apr 11, 2012 at 10:55:34AM +0200, Jean-Louis Durrieu wrote:
> I was thinking it would be a good idea to include in gmm.py such a
> mechanism.
Core sets are a very beautiful idea from the theoretical standpoint and
I'd love to have them in the scikit. We had even added them in the list
of i
haha okay, I'll give it a swing!
J
2012/4/11 Alexandre Gramfort
> > The simulations that are done in the last mentioned paper by Ming Yuan
> and
> > Yi Lin,
> > http://www2.isye.gatech.edu/statistics/papers/05-25.pdf, they find
> that the
> > NG seems to do
> > generally better than the LASSO
> The simulations that are done in the last mentioned paper by Ming Yuan and
> Yi Lin,
> http://www2.isye.gatech.edu/statistics/papers/05-25.pdf, they find that the
> NG seems to do
> generally better than the LASSO (figure 1)
if you can reproduce this figure using my gist I pay you a beer :)
I
@Olivier - no problem :)
@Alex
Appart from trying it ourselves to see how it fares, here're some findings:
The simulations that are done in the last mentioned paper by Ming Yuan and
Yi Lin,
http://www2.isye.gatech.edu/statistics/papers/05-25.pdf, they find that
the NG seems to do
generally bette
Le 11 avril 2012 11:53, Alexandre Gramfort
a écrit :
>
> it seems to me that it might be a good addition to the scikit if can
> convince ourselves with examples that it does better than a Lasso.
I agree.
Thanks for the write-up Jacques.
--
Olivier
http://twitter.com/ogrisel - http://github.com
> The algorithm proposed in this paper, is rather similar to that of the Lars
> LASSO, but with a complicating
> factor being a non-negative constraint on the shrinkage factor. (See eq. (2)
> in this paper)
> Once you've computed your shrinkage factor, you basically have your
> regression coeffic
Le 11 avril 2012 10:55, Jean-Louis Durrieu a écrit :
> Hi all,
>
> On Feb 7, 2012, at 8:47 AM, Olivier Grisel wrote:
>
>> 2012/2/6 Shishir Pandey :
>>>
>>> I am working with a dataset which too big to fit in the memory. Is there a
>>> way in scikits-learn to sub sample the existing dataset maintai
Here's a wee summary on the non-negative garrote (NG) i pieced together:
The original non-negative garrote from Breiman (1995) is basically a scaled
version of the least square estimate.
Basically take a OLS estimator and then shrink that estimator to obtain a
more sparse representation.
The shrin
Hi all,
On Feb 7, 2012, at 8:47 AM, Olivier Grisel wrote:
> 2012/2/6 Shishir Pandey :
>>
>> I am working with a dataset which too big to fit in the memory. Is there a
>> way in scikits-learn to sub sample the existing dataset maintaining its
>> properties so that I can load it in my RAM?
>
> We
13 matches
Mail list logo
