Re: [Scikit-learn-general] Tutorial on decision trees

2011-11-29 Thread Brian Holt
As a follow up, I found a description of the parallel tree training algorithm [2] that MSR used. Regards, Brian [2] http://budiu.info/work/budiu-biglearn11.pdf -- All the data continuously generated in your IT

[Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Alexandre Passos
Hi, Currently the fit function in GaussianProcess throws a weird exception when only one training example is passed to fit(): from sklearn.gaussian_process import GaussianProcess from sklearn.gaussian_process import GaussianProcess gp.fit([[1., 2.]], [-1.0]) gp.fit([[1., 2.]], [-1.0])

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Alexandre Gramfort
Hi Alex, I would say: if it makes sense to fit a GP with only one point: it should be fixed else: raise a nicer error message Alex On Tue, Nov 29, 2011 at 7:10 PM, Alexandre Passos alexandre...@gmail.com wrote: Hi, Currently the fit function in GaussianProcess throws a weird

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Alexandre Passos
On Tue, Nov 29, 2011 at 15:02, Alexandre Gramfort alexandre.gramf...@inria.fr wrote: Hi Alex, I would say: if it makes sense to fit a GP with only one point:    it should be fixed The thing is, I'm not quite sure. In the code I was writing it made sense to add points incrementally, but what

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Vlad Niculae
On Tue, Nov 29, 2011 at 10:02 PM, Alexandre Gramfort alexandre.gramf...@inria.fr wrote: Hi Alex, I would say: if it makes sense to fit a GP with only one point:    it should be fixed Note that even though it might not make any sense in practice, unless there's a mathematical reason that I'm

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Alexandre Gramfort
The thing is, I'm not quite sure. In the code I was writing it made sense to add points incrementally, but what you get out of a GP with only one point is something pretty silly, enough to be almost as useless as the output of a logistic regression with only one point. that would be my gut

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Kenneth C. Arnold
There is no maximum likelihood solution to a GP with a single training point, but you can certainly draw samples from the posterior; in fact, you can draw samples from the prior (without conditioning on data). That may help you determine if your covariance function is reasonable: samples from the

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Vincent Dubourg
Hi list, Indeed, I did not think about this usage of the GP predictor (actually I don't think DACE for Matlab handles this case either). In my opinion, fitting a GP with only one point does not make much sense even if it holds mathematically (i.e. you can compute the posterior distribution of

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Olivier Grisel
2011/11/29 Alexandre Passos alexandre...@gmail.com: On Tue, Nov 29, 2011 at 16:18, Vincent Dubourg vincent.dubo...@gmail.com wrote: @AlexP: What are you trying to do with this iterative construction? Are you trying to implement some optimization algorithm (like the efficient global optimizer

Re: [Scikit-learn-general] Issue with gaussian processes

2011-11-29 Thread Kenneth C. Arnold
On Tue, Nov 29, 2011 at 4:53 PM, Olivier Grisel olivier.gri...@ensta.org wrote: Now back to you problem I think we should support fitting models with just one sample just for the sake of consistency / continuity even if theds is no practical application of fitting models with a single sample: