Adding a formula interface with elm

[edwindj]

First of all, thanks for this package!

FWIW, I have added a formula interface with the function elm, which makes working with your package a bit smoother.
(tests are included). Feel free to adapt and include it in your package

Best,

Edwin de Jonge

elm

elm is a wrapper for elm_train, transforms a data.frame with formula into the necessary input for elm_train.
Furthermore, the returned object has some nice printing properties:

library(elmNNRcpp)
elm(Species ~ ., data = iris, nhid = 20, actfun="sig")

## Extreme learning model, elm (classification):
## 
## call:  elm(Species ~ ., data = iris, nhid = 20, actfun = "sig") 
## hidden units       : 20 
## activation function: sig 
## accuracy           : 0.9733333 
## 
## confusion matrix :
##             predicted
## observed     setosa versicolor virginica
##   setosa         50          0         0
##   versicolor      0         48         2
##   virginica       0          2        48

classification

Using predict on a elm object calls elm_predict with some extra's. For classification the type option
can be used returning either class, probabilities or raw

mod_elm <- elm(Species ~ ., data = iris, nhid = 20, actfun="sig")
# predict classes
predict(mod_elm, newdata = iris[1:3,-5])

## [1] setosa setosa setosa
## Levels: setosa versicolor virginica

# predict probabilities
predict(mod_elm, newdata = iris[1:3,-5], type="prob")

##         setosa versicolor   virginica
## [1,] 0.9775201 0.02021516 0.002264738
## [2,] 0.9519374 0.02871766 0.019344988
## [3,] 0.9620877 0.02854287 0.009369432

# predict elm output
predict(mod_elm, newdata = iris[1:3,-5], type="raw")

##        setosa  versicolor   virginica
## [1,] 1.018682 -0.02106638 0.002360102
## [2,] 1.009991 -0.03046901 0.020524740
## [3,] 1.020310 -0.03027020 0.009936443

data("ionosphere")
elm(class ~ ., data = ionosphere, nhid=20, actfun="relu")

## Extreme learning model, elm (classification):
## 
## call:  elm(class ~ ., data = ionosphere, nhid = 20, actfun = "relu") 
## hidden units       : 20 
## activation function: relu 
## accuracy           : 0.8603989 
## 
## confusion matrix :
##         predicted
## observed   b   g
##        b  89  37
##        g  12 213

regression

In regression the mse is returned.

data("Boston")
elm(medv ~ ., data = Boston, nhid = 40, actfun="relu")

## Extreme learning model, elm (regression):
## 
## call:  elm(medv ~ ., data = Boston, nhid = 40, actfun = "relu") 
## hidden units       : 40 
## activation function: relu 
## mse                : 20.20491

[mlampros]

@edwindj thank you for this PR.

[mlampros]

@edwindj I receive the following ERROR, WARNING and NOTEs (see the attached 00check.log file)
I checked the package locally because the Github Action was disabled. Can you have a look? I just enabled the Github Action, so if you make any commits the checks will be run

00check.log

(you can ignore the WARNING about the Insufficient package version)

[edwindj]

(teaching today, so won't be able able to look into it, before this evening)

[edwindj]

Looking at the github actions, it seems to be fixed 😀

[mlampros]

thank you very much @edwindj for this pull request. Give me a few days to update the CRAN version.
I'd like to add you also in the DESCRIPTION file as a contributor, is it ok by you?
I can use you existing Details

Edwin de Jonge, <edwindjonge@gmail.com>, ORCID: 0000-0002-6580-4718

Let me know. thanks again.

[edwindj]

You re welcome!
Ok, if you add me as a contributor, thanks!

Best,

Edwin

[mlampros]

@edwindj,

I intended today to submit the updated version to CRAN and I ran an R CMD check --as-cran on the cloned repository, however I receive the following error on Linux which comes from the examples (the tests seem ok),

* checking examples ... ERROR
Running examples in ‘elmNNRcpp-Ex.R’ failed
The error most likely occurred in:

> base::assign(".ptime", proc.time(), pos = "CheckExEnv")
> ### Name: elm
> ### Title: Fit an extreme learning model
> ### Aliases: elm
> 
> ### ** Examples
> 
> elm(Species ~ ., data = iris, nhid = 20, actfun="sig")
Extreme learning model, elm (classification):

call:  elm(Species ~ ., data = iris, nhid = 20, actfun = "sig") 
hidden units       : 20 
activation function: sig 
accuracy           : 0.9733333 

confusion matrix :
            predicted
observed     setosa versicolor virginica
  setosa         50          0         0
  versicolor      0         48         2
  virginica       0          2        48
> 
> mod_elm <- elm(Species ~ ., data = iris, nhid = 20, actfun="sig")
> 
> # predict classes
> predict(mod_elm, newdata = iris[1:3,-5])
Error in UseMethod("predict") : 
  no applicable method for 'predict' applied to an object of class "elm"
Calls: predict
Execution halted

can you reproduce this error locally using the cloned repository?

[edwindj]

Strange, it seems to work on my machine: git clone https://github.com/mlampros/elmNNRcpp cd elmNNRccp R CMD build . R CMD check --as-cran elmNNRcpp_1.0.3.tar.gz Furthermore, your github action: https://github.com/mlampros/elmNNRcpp/runs/3987209967?check_suite_focus=true seems to work for windows, linux and macos (examples do run correctly when clicking on the output on the right). So that's strange... Did you try `rhub::check_for_cran()` (only a maintainer is allowed to do that)? Best, Edwin Op za 23 okt. 2021 om 13:34 schreef Lampros Mouselimis < ***@***.***>:

…

@edwindj <https://github.com/edwindj>, I intended today to submit the updated version to CRAN and I ran an R CMD check --as-cran on the cloned repository, however I receive the following error on Linux which comes from the examples (the tests seem ok), * checking examples ... ERROR Running examples in ‘elmNNRcpp-Ex.R’ failed The error most likely occurred in: > base::assign(".ptime", proc.time(), pos = "CheckExEnv") > ### Name: elm > ### Title: Fit an extreme learning model > ### Aliases: elm > > ### ** Examples > > elm(Species ~ ., data = iris, nhid = 20, actfun="sig") Extreme learning model, elm (classification): call: elm(Species ~ ., data = iris, nhid = 20, actfun = "sig") hidden units : 20 activation function: sig accuracy : 0.9733333 confusion matrix : predicted observed setosa versicolor virginica setosa 50 0 0 versicolor 0 48 2 virginica 0 2 48 > > mod_elm <- elm(Species ~ ., data = iris, nhid = 20, actfun="sig") > > # predict classes > predict(mod_elm, newdata = iris[1:3,-5]) Error in UseMethod("predict") : no applicable method for 'predict' applied to an object of class "elm" Calls: predict Execution halted can you reproduce this error locally using the cloned repository? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#4 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAEEOHBNQVI2DTIYGR4ABY3UIKMTRANCNFSM5F7CLIEQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[mlampros]

@edwindj,

initially I thought the error comes from the git clone https://github.com/mlampros/elmNNRcpp command but it doesn't make any difference if I use git clone https://github.com/mlampros/elmNNRcpp.git

It seems that the error comes from the auto-generated .Rd files and especially of the Examples section. Once you clone the repository can you please delete all .Rd files of the 'man' directory in the elmNNRcpp package?

Then once deleted, the way I build and check a package that includes Rcpp code every time before I submit to CRAN is the following:

devtools::document(pkg = '/your_default_dir/elmNNRcpp')     # creates the updated .Rd files based on the .R files of the 'R' directory
remove_so_and_o_files = sapply(list.files('/your_default_dir/elmNNRcpp/src', pattern = ".o$|.so$", full.names = T), function(x) file.remove(x))
setwd('/your_default_dir/elmNNRcpp')
tools::package_native_routine_registration_skeleton(".", con = "/your_default_dir/elmNNRcpp/src/init.c", character_only = FALSE)
Rcpp::compileAttributes(verbose = T)
setwd('/your_default_dir')
system("R CMD build elmNNRcpp")
system("R CMD check --as-cran elmNNRcpp_1.0.3.tar.gz")
system("R CMD INSTALL elmNNRcpp_1.0.3.tar.gz")

I think that way you might reproduce the error compared to a simple R CMD build and check. The paths might need to be modified in the previous code snippet because they are adjusted for a unix OS

[mlampros]

moreover can you add

@importFrom stats model.frame model.matrix terms

in the elm() function

@importFrom stats model.frame model.matrix

in the predict.elm() function

and update the DESCRIPTION file by adding the 'stats' package?

Imports: Rcpp (>= 0.12.17), stats

[mlampros]

highly probable this error comes from the devtools::document() and the update of the NAMESPACE on Linux OS (I have a similar issue currently in another PR). Which OS do you use?

[mlampros]

I submitted the updated version to CRAN (1.0.4). This was actually a problem with the R version I used. After updating to 4.1.1 the issue was resolved.

[edwindj]

Great!