> On Jan 12, 2017, at 5:37 PM, Lorenzo Isella wrote:
>
> Dear All,
> I am fine tuning a Cubist model (see
> https://cran.r-project.org/web/packages/Cubist/index.html).
> I am a bit puzzled by its output. On a dataset which contains 275
> cases, I get non mutually exclusive rules.
> E.g., in the
I encountered this a few months ago and, in my case, the sentence had a noun
and verb but lacked a period at the end of the sentence. I tested that 'blah
blah blah.' would have passed in that version of R-devel.
Whenever I find a new rule or test with R CMD check, I tell myself that it must
be
> On Oct 5, 2014, at 4:51 PM, Lorenzo Isella wrote:
>
> Thanks a lot.
> At this point then I wonder: seen that my response consists of 5
> outcomes for each set of features, should I then train 5 different
> models (one for each of them)?
> Cheers
caret can only model one outcome at a time so
The data size isn't an issue. Can you send a reproducible example?
Max
On Jun 11, 2013, at 10:31 AM, Ferran Casarramona
wrote:
> Hello,
>
> I'm training a set of data with Caret package using an elastic net (glmnet).
> Most of the time train works ok, but when the data set grows in size I ge
Yes, I think the second link is a test build of a parallelized cv loop within
gbm().
On Mar 24, 2013, at 9:28 AM, "Lorenzo Isella" wrote:
> Thanks a lot for the quick answer.
> However, from what I see, the parallelization affects only the
> cross-validation part in the gbm interface (but it
train() in caret. See
http://caret.r-forge.r-project.org/
Also, the C5.0 function in the C50 is much more effective than J48.
Max
On Mar 23, 2013, at 2:57 PM, Nicolás Sánchez wrote:
> Good afternoon.
>
> I would like to know if there is any function in R to do LOOCV with these
> classi
caret has recursive feature and simple feature filters. I've got some genetic
algorithm code (using the GA package).
CORElearn also has the relief algorithm and a lot of different measures of
feature importance.
Max
On Mar 13, 2013, at 3:57 AM, "C.H." wrote:
> FSelector
>
> Maybe chi-sq i
What version of odfWeave and XML?
On Feb 18, 2013, at 11:49 AM, Paul Miller wrote:
> Hi Max,
>
> Sorry I didn't provide sufficient information. Below is my sessionInfo with
> all code included.
>
> Thanks,
>
> Paul
>
> sessionInfo()
> R version 2.13.1 (2011-07-08)
> Platform: x86_64-pc-
There is also C5.0 in the C50 package. It tends to have smaller trees that C4.5
and much smaller trees than J48 when there are factor predictors. Also, it has
an optional feature selection ("winnow") step that can be used.
Max
On Sep 21, 2012, at 2:18 AM, Achim Zeileis wrote:
> Hi,
>
> just
The train function in the caret package will do this. The trainControl function
would use method ="repeatedcv" and repeats = 100.
On Feb 18, 2012, at 2:15 PM, Greg Snow <538...@gmail.com> wrote:
> The validate function in the rms package can do cross validation of
> ols objects (ols is similar
This mostly happens when the data contain invalid column names (such as all
numbers). Try using make.names() on the datasets.
Max
On Oct 30, 2011, at 11:35 AM, Luisa Sêco wrote:
> Dear users,
>
> I'm using rpart for classification trees, but my code isn't working when I
> try to use all the
10 fold cv has high variation compared to other methods. Use repeated cv or the
bootstrap instead (both of which can be used with glmnet by way of the train()
function on the caret package).
Max
On Jul 23, 2011, at 11:43 AM, fongchun wrote:
> Hi Patrick,
>
> Thanks for the reply. I am ref
Xiaoqi,
You need to specify the sizes. There are other search algorithms that
auotmatically pick the size (such as genetic algorithms), but I don't have
those in the package yet.
Another approach is to use univariate filtering (see the sbf function in caret).
Max
On Mar 13, 2011, at 8:49 PM,
John,
What version of odfWeave and OO are you using?
Thanks,
Max
On Feb 22, 2011, at 3:17 PM, "Prof. John C Nash" wrote:
> Using R2.12.1 on Ubuntu 10.04.1 I've tried to run the following code chunk in
> odfWeave
>
> <>=
> x<-seq(1:100)/10
> y<-sin(cos(x/pi))
> imageDefs <- getImageDefs()
>
If you want to get honest estimates of accuracy, you should repeat the feature
selection within the resampling (not the test set). You will get different
lists each time, but that's the point. Right now you are not capturing that
uncertainty which is why the oob and test set results differ so mu
If you do use correlation, you should think about doing it on the log or sort
scale.
The train() function in the caret package can estimate performance using
resampling. There are examples in ?train that show how to define custom
performance measures (I think it shows how to do this with MAD es
You should try different tuning parameters; the defaults are not likely to work
for many datasets. I don't use the polynomial kernel too much but scale
parameter values that are really of could cause this. Unlike the rbf, I don't
know of any good techniques for estimating this.
Max
On Jan 25,
Greg and Frank,
Thanks for the replies. I didn't express myself very well; I'm not interest in
the model fitting aspect. I'd just like to get the full set of dummy variables
(optimally from model.matrix)
Max
On Dec 6, 2010, at 10:29 PM, Frank Harrell wrote:
>
> Given a non-singular fit, the
> is ipred:errorest method is good enough for validating (or cross check) my
> svm result?
Yes, if you know what values of the tuning parameters to use, but I don't know
why it was failing.
Max
__
R-help@r-project.org mailing list
https://stat.ethz.c
I think the issue is that you really can't use the training set to judge this
(without resampling).
For example, k nearest neighbors are not known to over fit, but a 1nn model
will always perfectly predict the training data.
Max
On Oct 23, 2010, at 9:05 AM, "Liaw, Andy" wrote:
> What Breim
If you use the index argument of the trainControl() function in the caret
package, the train() function can be used for this type of resampling (and
you'll get some decent summaries and visualizations to boot)
Max
On Jul 21, 2010, at 7:11 AM, "Tim Howard" wrote:
> Coll,
>
> An alternative ap
21 matches
Mail list logo
