[Scikit-learn-general] Understanding Sparsity Example

Hi, regarding the example http://scikit-learn.org/dev/auto_examples/linear_model/plot_ols_3d.html I have the following question: The text says, feature 2 has a strong coefficient. How is this illustrated? In the first plot (Y-X_2) I would say the plane/line is almost horizontal, which means n

Re: [Scikit-learn-general] Scikit-learn tutorial at Scipy 2014 (deadline: March 14)

I am sorry. I, for one, will not be going there. I cannot afford to go to scipy every year. Gaël On Mon, Mar 03, 2014 at 12:25:53PM -0800, Jacob Vanderplas wrote: > Hi folks, > It's that time of year again: proposal submissions for the Scipy 2014 > conference are open, and due March 14! You can

[Scikit-learn-general] Scikit-learn tutorial at Scipy 2014 (deadline: March 14)

Hi folks, It's that time of year again: proposal submissions for the Scipy 2014 conference are open, and due March 14! You can read about it here: https://conference.scipy.org/scipy2014/about/ I'm serving as co-chair of the tutorials this year, and we're excited to have three tracks of tutorials

Re: [Scikit-learn-general] online/minibatch multilabel classification

Hey, Just so that my silence isn't misinterpreted: I am OK either way, and I trust your judgement (or that of other core devs well-versed in these topics). Gaël On Mon, Mar 03, 2014 at 04:07:05PM +0100, Lars Buitinck wrote: > Dear all (esp. pprett, ogrisel), > For a largeish multilabel classif

Re: [Scikit-learn-general] WardAgglomerative modify inertia criteria

The implementation trick in the Ward algorithm is that only the first and the second moment of the clusters (in other terms their mean and variance) need to be kept to be able to recompute the inertia criteria. I don't have the code under hand right now, but I would believe that m_1 and m_2 are the

[Scikit-learn-general] online/minibatch multilabel classification

Dear all (esp. pprett, ogrisel), For a largeish multilabel classification experiment (600k samples) I need online learning: not only is the set large, but I also want to simulate the streaming in of samples over time. The samples are timestamped document snippets collected over a period of years.

[Scikit-learn-general] WardAgglomerative modify inertia criteria

Hi, I am working on a clustering algorithm and want to modify the inertia crieria for WardAgglomerative function. As far as understand it takes place at _hierarchical.pyx file in compute_ward_dist function. I can understand that the following line (m_1[row] * m_1[col]) / (m_1[row] + m_1[col]) co