Bryan:
>
> Hi,
>
> I was wondering if Statdataml is currently the preferred way to
> represent statistical data in XML in R.
This is hard to tell. We think it is a _possible_ way doing that. Do you
have some alternatives in mind?
And also if the Statdataml api
> provides ways to load the XML
Hi all,
there is tune() in the e1071 package for doing this in general, and,
among others, a tune.nnet() wrapper (see ?tune):
> tmodel = tune.nnet(Species ~ ., data = iris, size = 1:5)
> summary(tmodel)
Parameter tuning of `nnet':
- sampling method: 10-fold cross validation
- best paramete
ness from a specified property (e.g., symmetry,
transitivity, etc.).
David and Kurt.
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Tel: +43-1-313 36 4393
Fax: +43
Chris:
yes, this is indeed a bug (in predict.svm) - will be fixed in the next
release of e1071.
Thanks for pointing this out,
David
--
Hello all,
I am trying to use the "svm" method provided by e1071 (Version: 1.5-16)
together with a matrix provided by the Sparse
Dear useRs,
a new package for computing distance and similarity matrices made it to
CRAN, and will propagate to the mirrors soon.
It includes an enhanced version of "dist()" with support for more than
40 popular similarity and distance measures, both for auto- and
cross-distances. Some importa
Adschai:
The function is written in C++, so debugging the source code of the R
svm() function will not really help. What you can do to make the C-code
more verbose is the following:
- get the sources of e1071
- in the src/ directory, look up the svm.cpp file
- In line 37, there is:
#if 0
void
Adschai:
here is an example for class.weights (isn't it on the help page?):
data(iris)
i2 <- iris
levels(i2$Species)[3] <- "versicolor"
summary(i2$Species)
wts <- 100 / table(i2$Species)
wts
m <- svm(Species ~ ., data = i2, class.weights = wts)
Cheers,
D
Dear useRs,
it is our great pleasure to announce the new package "relations" to
appear on all CRAN-mirrors soon.
This package provides data structures and methods for creating and
manipulating relations, relation ensembles, sets, and tuples. The
feature list includes:
* creation of relations by
It's in package cba (sdists()).
David
--
I have a column of words, for example
"DOG"
"DOOG"
"GOD"
"GOOD"
"DOOR"
...
and I am interested in creating a matrix that contains the string
edit distances between each pair of words. I am this close -> ' '
<- to writing the algorit
Oldrich:
>
> The columns in data passed to svm need to contain only numeral values.
This is not correct, svm() of course also accepts factors and then
builds a model matrix similar to lm(). But it won't accept, e.g.,
character vectors.
> I simply assigned a number to each category of each fea
> Hello:
> > >
> > > Does anyone know if there exists a package that handles methods for [
> for
> > > dist objects?
> > >
> > > I would like to access a dist object using matrix notation
> > >
> > > e.g.
> > >
> > > dMat = dist(x)
> > > dMat[i,j]
You can use the [[ operator defined f
ea is to create a package for that, since XMLA and JOLAP
> specifications
> > > should able us to do so !
> > >
> > >
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wi
Hello (whoever you are),
your data looks problematic. What does
head(ne_span_data)
reveal?
BTW, svm() will not handle NA values.
Best
David
-
Hello. I'm new to R and I'm trying to solve a classification problem. I have
a training dataset of about 40,000 rows and 50 column
David:
>I ‘m trying to draw ONE table that summarize SEVERAL categorical
>variables
>according to one classification variable, say “sex”. The result would
>look
>like several contingency tables appended one to the other. All the
>variables
>belong to a data frame.
>The summary.formula
Aimin:
The problem is that the columns you choose for training (only 4
variables) do not match the ones used for prediction (all except y).
David
I try to use naiveBayes
> p.nb.90<-naiveBayes(y~aa_three+bas+bcu+aa_ss,data=training)
>
pr.nb.90<-table(predict(p.nb.90,training[,-13],ty
e my code. it doesn't work.
> It is pity not saving my original code.
>
> Now the question is the plot I get using your code is different from
> what I got before.
> Moreover I did remember I use plot(m.svm,p5.new,As~Cur)
>
> Do you know why?
>
> Thanks,
>
>
.new$Y) <- list(Out=c(1), In=c(0))
>> attach(p5.new)
>> m.svm<-svm(Y~P+Aa+As+Cur,data=p5.new)
>> summary(m.svm)
>
> Call:
> svm(formula = Y ~ P + Aa + As + Cur, data = p5.new)
>
>
> Parameters:
>SVM-Type: C-classification
> SVM-Kernel: radial
>
t;null"), mar[2], legend_width) :
>>>> It is invalid to combine unit objects with other types
>>> Works for me. Which version of R is this? R-2.4.0, R-patched or
>>> R-devel?
>>>
>>
>
>
>
--
Dr. David Meyer
Department of Information
Vince:
the implementations for both are different, so this might happen
(although undesirably).
Can you provide me an example with data (off-list)?
David
__
R-help@stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do r
testing whether the 2 odds ratios are the same?
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Tel: +43-1-313 36 4393
Fax: +43-1-313 36 90 4393
HP: http://wi.wu-wien.ac.at
model (svm from the package e1071) but
I have not succeed.
Does anyone knows a way to get that value?
Pau
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Tel: +43-1-313 36 4393
;TOF" "EXT" "Acode"
> dim(new.dat)
[1] 175 3
Now try to predict:
> predict(cd1.svm,newdata=new.dat)
Error in "names<-.default"(`*tmp*`, value = c("1", "2", "3", "4", "5", :
'
Dieter,
there is no way of fixing the range of the residuals yet, we will add sth.
like a ylim argument to legend_foo().
Thanks for pointing this out,
David
PS: there is no mosaicplot() function in vcd, but a mosaic() ...
--
Dr. David Meyer
Department of Information Systems and Operations
TRUE when training the model, but doesn't show an
> example. Then I try to predict with probabilities, I still only get
> classifications back. Anyone get this to work and can help me out?
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economi
r plot.svm, which has a
simple example (that works)
## a simple example
library(MASS)
data(cats)
m <- svm(Sex~., data = cats)
plot(m, cats)
I don't see what the difference between their example and mine.
--
Dr. David Meyer
Department of Information Systems and Operat
t;repetitive:
>How can one (elegantly) calculate the distance between data points (in the
>transformed space, I suppose) and the hyperplane that separates the 2
>categories when using svm() from the e1071 library?
>thanks a lot,
>Hans
--
Dr. David Meyer
Department of Informa
rror on some objects but I just was not so lucky to
> > meet one?.
> >
> > W2000, R2.2.0, vcd package Built: R 2.2.0; ; 2005-11-
> > 22 14:23:44; windows,
> >
> > Best regards.
> >
> > Petr
> >
> > Petr Pikal
> > [EMAIL PROTECTED]
&
table(), etc.
are available
* New panel function pairs_diagonal_text() for pairs()
* The alternative legend function legend_fixed() now looks more similar
to the legend of mosaicplot() in base R
Comments are more then welcome!
David, Achim, Kurt.
PS: If you like modern art, try out demo(mondrian
Well, since knn() can't handle incomplete data as it says, you can
choose to either omit incomplete observations (e.g., using na.omit()),
or to impute the data if the conditions are met (missingness at random,
...); see, e.g., packages cat, mix, norm, and e1071 for that.
HTH,
David
l yet for
time series.
Cheers,
David
> Its possible that what works already is enough for Amir. For example,
>
> library(e1071)
> library(dyn)
> set.seed(1)
> y <- ts(rnorm(100))
> y.svm <- dyn$svm(y ~ lag(y))
> yp <- predict(y.svm)
> ts.plot(y, yp, col = 1:
Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Fax: +43-1-313 36x7
intend to set 1 for class one, 49 for class 2. How do I
do that? Just weights=49?
Meanwhile I'd like to try svm(e1071), again, how do I setup
"class.weights"? Thanks.
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economics and Business
y
table.
Happy drawing!
David, Achim, Kurt
--
Dr. David Meyer
Department of Information Systems and Operations
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Fax: +43-1-313 36x746
Tel: +43-1-313 36x4393
HP: http://wi.wu-wien.ac.at/~meyer/
>
> Dear David, Dear Friends,
>
> After any running svm I receive different results of Error estimation
> of 'svm' using 10-fold cross validation.
using tune.svm(), or the `cross' parameter of svm()?
> What is the reason ? It is caused by the algorithm, libsvm , e1071 or
> something els?
ith the math behind svms, so my assumtions maybe
> completely wrong ... or obvious. Will the svm automatically expand the
> factors into a binary matrix?
yes.
> If I add numeric variables outside the range of 0 to 1 do I have to
> scale them to have 0 to 1 range?
svm() will scale your data
Vivek,
I certainly would agree that every help page, including the one of
svm(), could be improved, but I think it is not _that_ deficient. In
particular, it tells you which parameters are used in the various
kernels available.
Have you read the corresponding article in R News (basically contain
Stephen:
you need to supply the parameter ranges, your call did not tune anything
at all.
best.svm() is really just a wrapper for tune.svm(...)$best.model. The
help page for 'tune()' will tell you more on the available options.
HTH,
David
[...]
> svm.model = best.svm(data[1:3000,1:23],data[1:
Support
>Vectors , gamma, cost , nu , >epsilon , after model<- tune.svm( ) ?
>these are not possible? I receive only "Error estimation of 'svm' "
>with model and summary(model) functions.
Clear from the above, I think.
HTH,
David
>Best Wishes and so man
> Thank you for your answer,
> but my problem concerns the support vectors. Indeed the two classes
> are well separated and the hyperplane is linear but the support
> vectors aren't aligned in parallel to the hyperplane. And according to
> me, the support vectors (for each class) should be aligned
w and threshold b ?
On the data set, and therefore also on w and b.
Best,
David
--
Dr. David Meyer
Department of Information Systems and Process Management
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Fax:
ALERT!
This e-mail, in its original form, contained one or more attached files that
were infected with a virus, worm, or other type of security threat. This e-mail
was sent from a Road Runner IP address. As part of our continuing initiative to
stop the spread of malicious viruses, Road Runner s
Joao:
1) The error message you get when setting nu=0 is due to the fact that
no support vectors can be found with that extreme restriction, and this
confuses the predict function (try svm(, fitted = false): the model
returned is empty). In fact, the C++ code interfaced by svm() clearly
allows
This is a bug, thanks for pointing this out.
Fixed for the next release of e1071.
David
-
Hello,
I would like to read and write sparse matrices using the
functions write.matrix.csr() and read.matrix.csr()
of the package e1071. Writing is OK but reading back the
matrix fails:
Raj:
The references given on the help page will tell you.
Best,
David
-
Hi All,
In package e1071 for SVM based classification, one can get a probability
measure for each prediction. I like to know what is method that is used
for
calculating this probability. Is it calculated using logi
ssage: "Die Anweisung in 0x6c48174d verweist auf Speicher
> 0x. Der Vorgang "read" konnte nicht auf dem Speicher
> ausgef_hrt werden. ."
Which version of e1071 are you using?
There has been a memory leak problem until 1.5-1 which could very well
cause this
> I updated the e1071 package but still can't find
> the other three arguments for plot.svm.
It's in e1071 since version 1.5-3. (current version: 1.5-4).
> In addition, I can plot a
> gray-colored contour region by adding the argument "col = c(gray(0.2),
> gray(0.8))". But I failed to change th
ds) for
> training. The error is in the svm function. May be because SVM uses
> other functions, but it happens when I run svm.
>
> Thanks a lot for the help
>
> Joao
> ___
> FEUP - Engineering Faculty, Porto University
> Engineering and Industri
)?
Thanks for any help
Joao
--
Dr. David Meyer
Department of Information Systems
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Fax: +43-1-313 36x746
Tel: +43-1-313 36x4393
HP: http://wi.wu-wien.ac.at/~meyer/
___
?
see filled.contour(). The input is determined by a grid of predicted
values.
Best,
-d
> Thank you very much,
> Frank
--
Dr. David Meyer
Department of Information Systems
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Fax: +43-1-313
round this?
--
Dr. David Meyer
Department of Information Systems
Vienna University of Economics and Business Administration
Augasse 2-6, A-1090 Wien, Austria, Europe
Fax: +43-1-313 36x746
Tel: +43-1-313 36x4393
HP: http://wi.wu-wien.ac.at/~meyer/
__
[EMAIL
Uwe:
[the language of the list is English!]
Try using a *factor* for classification. The described behavior
(segfault when using class.weights with a *numeric* dependent variable)
should be fixed in the current version of e1071 (1.5-2), so please check
if you are using the latest version of e107
d <= .5] = 0
> table(testing$similarity, pred)
pred
0 1
0 8 30
1 26 107
>
and then taking out the kappa statistic to see if I am getting anything
significant.
I get kappas of 15 - 17% - I don't think that is very good. I know
kappa is really for comparing the
> Content-Type: text/plain; charset=ISO-8859-1
>
>
> Does anyone knows what is the nu-regression option for the type
> parameter in svm (from package e1071)? I cannot find any explanation
> on that and I have a reasonable understanding on svm fundamentals.
>
> Thanks
>
>
gested, you might want to check if
your target variable is indeed a factor. In case it is and you still get
values greater than 1, drop me a mail with a piece of code (and data)
enabling me to reproduce the phenomenon.
Best,
David
--
Dr. David Meyer
Department of Information Systems
Vienna Univers
Have you considered the read.matrix.csr() function in pkg. e1071? It
uses another sparse input format, but perhaps you can easily transform
your data in the supported one. Also, in my experience, data frames are
not the best basis for a sparse format since they might turn out to be
very memory cons
cation results per
>cross >validation out the svm? I mean e.g. numbers about the true
>positives ,fp,fn,tf ?
No, because the accuracies are not computed using a confusion matrix.
The computation is internally done in C.
>How do I get a list of the classified examples ?
--
Dr. David M
You could look at
@Article{ e1071-papers:meyer+leisch+hornik:2003,
author= {David Meyer and Friedrich Leisch and Kurt Hornik},
title = {The Support Vector Machine under Test},
journal = {Neurocomputing},
year = 2003,
month = {September
You might consider using the `weight' argument of svm().
Best,
David.
Hi!
I am using R 1.8.1 and the svm of the e1071 package for classification.
The problem is that I have unbalanced classes e.g. the first one is much
bigger than the second one and therfore the svm is biased to the first
clas
On Mon, 3 Nov 2003 [EMAIL PROTECTED] wrote:
> I am trying to understand what is the difference between linear and
> polynomial kernel:
>
> linear: u'*v
>
> polynomial: (gamma*u'*v + coef0)^degree
>
> It would seem that polynomial kernel with gamma = 1; coef0 = 0 and degre
On Fri, 31 Oct 2003 [EMAIL PROTECTED] wrote:
> > rng <- list(gamma = 2^(-1:1), cost = 2^(2:4))
> > rng
> $gamma
> [1] 0.5 1.0 2.0
>
> $cost
> [1] 4 8 16
>
> > obj <- tune.svm(pIC50 ~ ., data = data, ranges = rng)
> Error in tune(svm, train.x = x, data = data, ranges = ranges, ...) :
>
> This suggests to me that data are scrambled each time - the last time I
> looked at libsvm python interface
> this is what was done. Is this the same here (I hope) ?
yes.
g.,
David
__
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mail
> > If I try the model below, R seems to grumble with a complaint.
> >
> > glm(cbind(Y,1-Y) ~ X, family = binomial)
> >
> > non-integer counts in a binomial glm! in: eval(expr, envir, enclos)
> >
For binomial models (as described in the help page), the response must be
either a factor or a n x 2 m
> Is there an alternative way of "dispatching" the printing, such that
> the usual print method for loglm is used after doing what is special
> for hllm?
You might want to have a look at `NextMethod()'
Best,
David
__
[EMAIL PROTECTED] mailing list
https:
rt. I will change the code to give a better error message in
such cases.
Best,
David.
On 2003.09.03 18:32, Luis Miguel Almeida da Silva wrote:
The data goes in attachment. I used ts to create data.ts
data.ts<-ts(data=data,start=c(2001,1),frequency=12)
-Original Message-
From: Dav
How did you construct `data.ts'? Can you send me the file?
best,
David
On 2003.09.03 15:57, Luis Miguel Almeida da Silva wrote:
Dear helpers
I'm having a problem with function HoltWinters from package ts. I have
a time series that I want to fit an Holt-Winters model and make
predictions for the n
And in package vcd function ternaryplot().
g.,
-d
On 2003.07.15 09:33, Tobias Verbeke wrote:
> Are there enough geochemists using R already that he'd find
> like-minded people to discuss technical issues with if he _did_
> switch to R? Is there a package somewhere already that does ternary
> and o
Andrew,
1) The current R version is 1.7.1
2) Which version of `e1071' are you using?
3) Does the `e1071.so' file exist (in e1071/libs)?
best,
David.
On 2003.06.24 21:43, Andrew Perrin wrote:
After upgrading to 1.7.0 under debian linux, I can't get e1071 working
properly.
The first problem I had wa
On 2003.06.13 02:11, g wrote:
Hi,
Can someone set me straight as to how to write formulas in R to
indicate:
complete independence [A][B][C]
Freq ~ A + B + C
joint independence [AB][C]
Freq ~ A * B + C
conditional independence [AC][BC]
Freq ~ A * C + B * C
nway interaction [AB][AC][BC
Support Vector Machines
Description:
`svm' is used to train a support vector machine. It can be used
to
carry out general regression and classification (of nu and
epsilon-type), as well as density-estimation. A formula interface
is provided.
David Meyer <[EMAIL PROTECTED]>
On 2003.06.12 09:15, Uwe Ligges wrote:
Iouri Tipenko wrote:
Dear R-Users,
I'm a master student in Mathematics and Statistics at Carleton
University, Ottawa, Canada.
I'm studying Clustering methods including different related
algorithms. One of them is Support Vector Clustering algorithm.
I was w
>
> I am trying to compile the package e1071 (version 1.3-11) with R CMD
> INSTALL. I tried with R 1.7.0 on Redhat Linux 2.4.7-10 and R 1.6.2 on
> Linux 2.4.9-34smp but keep getting the same error message during
> configure :
>
> WARNING: g++ 2.96 cannot reliably be used with this package. Please
Try `goodfit' in package `vcd'.
g.,
-d
Mag. David MeyerWiedner Hauptstrasse 8-10
Vienna University of Technology A-1040 Vienna/AUSTRIA
Department of Tel.: (+431) 58801/10772
Statistics and Probability Theory Fax.: (+431) 58801/10798
xcel problem or an sql problem.? I did not find anything in
> >> the r-help archives relative to this problem.
> >> Thanks for any help
> >>
> >> __
> >> [EMAIL PROTECTED] mailing list
> >> https:
E.g., in package e1071.
best,
-d
Mag. David MeyerWiedner Hauptstrasse 8-10
Vienna University of Technology A-1040 Vienna/AUSTRIA
Department of Tel.: (+431) 58801/10772
Statistics and Probability Theory Fax.: (+431) 58801/10798
On Sat, 8 Mar
tp://stat.ethz.ch/~hennig/
> [EMAIL PROTECTED], http://www.math.uni-hamburg.de/home/hennig/
> ###
> ich empfehle www.boag.de
>
> __
> [EMAIL PROTECTED] mailing list
> http://www.stat.math.ethz.ch/
-8013, fax -8007
> :: [EMAIL PROTECTED]
> :: http://www.informatik.uni-freiburg.de/~helma/
>
> __
> [EMAIL PROTECTED] mailing list
> http://www.stat.math.ethz.ch/mailman/listinfo/r-help
--
Mag. David MeyerWiedner Hauptstrasse 8-10
Vienna University of Techn
76 matches
Mail list logo