Hi David
I did try Ridge Regression as per my original message, but didn't get
any results. Maybe I'm implementing it incorrecly.
Generally, the data set should work fine I think. I've correlated the
features to the dependent variable and get high correlations for them
independently. I'm not s
On Thu, Aug 9, 2012 at 2:08 PM, Zach Bastick wrote:
> But as you can see, the predictions are absolutely terrible, no matter
> what I do.
> The training set predictions are quite accurate though. From my reading,
> this could be due to over fitting. However, I don’t see how simple
> linear model
On 10 August 2012 01:53, mathieu lacage wrote:
> hi,
>
> I have been using sklearn for a while now but I only recently started to
> figure out how to make sure I am using it correctly and that the results I
> get are meaningful so, the following questions are fairly general questions
> about mach
Thanks
I've created a script here with the two datasets (one for training the
model and one for testing the model), and a loader file to get the data
into python variables.
https://gist.github.com/3309139
I can't seem to get anything out of the data, I'd really appreciate the
help figuring ou
Hi GSOCers (Vlad and Immanuel),
The GSOC is getting to an end, and we are in the last rush. It's really a
pitty, as the projects seem to be entering a super-productive phase in
which pull request with significant speed ups are popping up.
As we want to benefit as much as possible of the remaining
Hi Zach,
if you provide a gist with your evaluation setup (similar to this one
[1]) I can look into it.
best,
Peter
[1] https://gist.github.com/3266657
2012/8/9 Zach Bastick :
> I’m having some conceptual trouble with this supervised machine learning
> project (regression) that hopefully someo
I’m having some conceptual trouble with this supervised machine learning
project (regression) that hopefully someone can help me with.
I am trying to do sentiment analysis on texts (scoring them from -10 to
+10) based on a human-scored training set.
Training set:
Cases = 35
Score Mean = 0.77
Sc
hi,
I have been using sklearn for a while now but I only recently started to
figure out how to make sure I am using it correctly and that the results I
get are meaningful so, the following questions are fairly general questions
about machine learning applied to text content. Hopefully, someone who
As Alex is not wearing the cheerleader outfit, I think I won't be able to make
it ;)
Andy
- Ursprüngliche Mail -
Von: "Alexandre Gramfort"
An: [email protected]
Gesendet: Donnerstag, 9. August 2012 14:58:37
Betreff: Re: [Scikit-learn-general] Sprints at EuroScip
On Thu, Aug 9, 2012 at 3:28 PM, Vlad Niculae wrote:
> Andy, Mathieu:
> The docs are lacking guidelines and examples on how to tune SVR
> parameters. IIUC, C, gamma, etc should be use just as in SVC. The tricky
> part is epsilon, how should it be set? What are sensible defaults and a
> sensible gr
On Thu, Aug 9, 2012 at 1:30 PM, Andreas Müller wrote:
> Sorry for being unspecific.
> Using the kernel should be more efficient with higher degree polynomials
> and when having
> many features. The dimensionality of the explicit features grows very fast
> with the degree while the cost
> of the ke
I won't be able to attend the sprints either...
Otherwise I would have for sure volunteered to wear this cheerleader outfit :)
Alex
On Wed, Aug 8, 2012 at 6:43 PM, Gael Varoquaux
wrote:
> Hey list,
>
> The euroscipy organizers are asking me to organize a bit the sprint. I
> have been trying to i
Dear JP,
Firstly, sorry for the delayed reply..
You appear to be on to something. I played around with your code and the
classifier and after discussing it with @Gael, it seems clear that there's
a pig in the truffle-patch here.
It seems that in the case you sample code, the problem occurs when
Andy, Mathieu:
The docs are lacking guidelines and examples on how to tune SVR parameters.
IIUC, C, gamma, etc should be use just as in SVC. The tricky part is epsilon,
how should it be set? What are sensible defaults and a sensible grid search
range?
Thanks,
Vlad
On Aug 9, 2012, at 13:30 , An
Hey Paolo.
Sorry for being unspecific.
Using the kernel should be more efficient with higher degree polynomials and
when having
many features. The dimensionality of the explicit features grows very fast with
the degree while the cost
of the kernel computation stays the same.
Also SVMs work quite
Hi Andy,
On Thu, Aug 9, 2012 at 11:53 AM, Andreas Müller wrote:
> Also you might need to normalize the data and set the value of C.
> Still this should work better than doing the explicit expansion.
>
What do you mean exactly by work better?
Paolo
---
> Please review:
> https://github.com/scikit-learn/scikit-learn/pull/1003
> I think that I made sure that pretty much every estimator was
> well-behaved. A lot of small changes. This can PR benefit from many
> eyes.
>
Woah that was fast. Thanks!
> That common testing framework is a pleasure, And
Also you might need to normalize the data and set the value of C.
Still this should work better than doing the explicit expansion.
- Ursprüngliche Mail -
Von: "Mathieu Blondel"
An: [email protected]
Gesendet: Donnerstag, 9. August 2012 09:53:18
Betreff: Re: [Sciki
Alternatively you could look at the output of "decision_function" in LinearSVC.
These do not represent probabilities, though.
Andy
- Ursprüngliche Mail -
Von: "Gael Varoquaux"
An: [email protected]
Gesendet: Donnerstag, 9. August 2012 05:50:14
Betreff: Re: [Scik
On Thu, Aug 9, 2012 at 4:02 PM, Zach Bastick wrote:
> I'm going to manually stop it now by closing the python window. Am I
> doing something wrong?
>
>
It probably means that epsilon is not well tuned. You can try
SVR(kernel="linear") to see how it fares compared to least squares.
Mathieu
--
I ran:
>> model = SVR(kernel="poly", degree=2)
but the % Error of the prediction is worse than using simple Ordinary
Least Squares using:
>> linear_model.LinearRegression()
It's also much slower. I changed the degree to 4 to see if the results
of the prediction got any better, but it's taking
21 matches
Mail list logo
