NEWS

---

NNbenchmark 3.2.1

• start parallelization for trainPredict_1pkg() (inspired from boot pkg)
• add extra-argument lib.loc in trainPredict_1pkg() and trainPredict_1data()

NNbenchmark 3.2.0 (2021-06-05)

• NNpredict.R:
• trainPredict_1mth1data(..., odir = "."). Before was "~/" which is incorrect
• trainPredict_1data(..., odir = ".") (as path have no final slash in R)
• trainPredict_1pkg(..., odir = ".")
• trainPredict_1mth1data() file.path(odir, ...) replaces paste0(odir, ...)

NNbenchmark 3.1.9 (2021-03-09)

• All URLs tested with the new package urlchecker_1.0.0 published on 2021-03-04
• datasets.R: http replaced by https
• datasets.R: NIST pages replaced by a redirection from InModelia's website
• DESCRIPTION and a_intro.R: New links to theairbend3r, Github and InModelia
• DESCRIPTION: Akshaj and Salsabila Orcid numbers

NNbenchmark 3.1.6 (2020-09-11)

• v_3.1.5 erased since mixed with 2019 code
• v_3.1.6 copied from v_3.1.4

NNbenchmark 3.1.4

• timediff() updated time0 <- NA ; time0 <<- time0 detected by CRAN
• Authors@R modified: correct PK email
• Author: orcid number for PK and CD

NNbenchmark 3.1.3

• timediff() updated time0 <<- NA
• NNSummary() and NNtrainPredict() cat(), print() modified

NNbenchmark 3.1.1

• DESCRIPTION: add Authors@R
• clearNN() and detachNN() further simplified

NNbenchmark 3.1.0

• change version number and sent to CRAN

NNbenchmark 3.0.7

• simplify clearNN()
• Christophe Dutang added as author
• Email address: rpackages--@--inmodelia
• false link replaced by the table header "Packages" in README.md
• NNtrainPredict.R: odir <- tempdir() rather than getwd()

NNbenchmark 3.0.5

• allow for nrep = 1
• added examples to NNtrainPredict
• changed echoreport to logical

NNbenchmark 3.0.4

• remove commented traces
• add an argument to trainPredict_1mth1data()
• add mWoodG1 rda file (Simon Wood's GAM test file for Gaussian family)

NNbenchmark 3.0.3

• NNtrainPredict now has the options for csv outputs, same as 2019 legacy code
• round time objects in NNsummary

NNbenchmark 3.0.2

• adjusted NNtrainPredict and NNsummary functions to new timing functions

NNbenchmark 3.0.1

• DESCRIPTION: Roxygen: list(old_usage = TRUE)
• timestart(), timediff()

NNbenchmark 3.0.0

• NNTrain_Predict.R for NEW 2020 code do.call() implementation to train and predict NNs
• Move NNbenchmark2020 to top directory. This is new NNbenchmark
• Move NNbenchmark2019 to a new directory
• Fix the date in DESCRIPTION

NNbenchmark 2.3.7

• Added MSE and WAE to funRMSE.R
• Made a new R file: NNSummary.R for 2020 summary function

NNbenchmark 2.3.6

• ht() and ht9() fully rewritten. Accept now array and data.frame with matrices inside
• README.md added

NNbenchmark 2.3.5

• DESCRIPTION and a_intro.R: add the html links to GitHub and AkshajVerma.com

NNbenchmark 2.3.4

• DESCRIPTION: Modify the title to match title case (recommended by CRAN)

NNbenchmark 2.3.3

• DESCRIPTION: Modify the title
• All files: Add value to Rd files (return in roxygen comments) for every functions

NNbenchmark 2.3.2

• prepareZZ: extend donttest to run examples in less than 5 seconds

NNbenchmark 2.3.1

• add2csv: examples rewritten with tempfile()
• datalist rewritten with format (file_without_ext: obj1, ..., objn)
• use for this purpose the hidden function tools:::list_data_in_pkg()

NNbenchmark 2.3

• NNdatasets: Change uNeuroOne from 3 to 2 neurons and 10 to 7 parameters
• NNdatasets: mRef153: correct a typo in the number of neurons (from 4 to 3)
• u+mNNdatasets: Update the documentation with the number of parameters for each dataset
• prepareZZ: rearrange the examples with donttest to run less than 5 seconds
• DESCRIPTION: Yifu Yan as author of the timeR function + class (timeR package)
• Test CRAN check and prepare the upload to CRAN

NNbenchmark 2.2

• NNdataSummary()

NNbenchmark 2.1.1

• clearNN() donotdetach now includes kableExtra, dplyr, stringr

NNbenchmark 2.1

• detachNN() (to move library(pkg) out of the loop in NNTemplate)

NNbenchmark 2.0.1

• Date 2019-08-04 added at the first line of *.R files
• Linked to GitHub. Package is rebuilt before commit (to force change in NAMESPACE)

NNbenchmark 2.0

• column 'params' added to the timer table. Columns in the data.frame are:
• timer$stop(event, start, end, duration, RMSE, MAE, stars, params, comment)
• Connect to GitHub

NNbenchmark 1.9

• New disk
• timer$stop(event, RMSE, MAE, stars, comment, printmsg) column 'stars' added to the table

NNbenchmark 1.8.6

• timer$stop(printmsg = TRUE) Use printmsg = FALSE to avoid printing

NNbenchmark 1.8.2

• clearNN(): ZZ is (now) detached from the environments
• clearNN(): donotdetach now indludes packages rmarkdown, knitr and captioner
• timer$stop suppress the message on the time elapsed

NNbenchmark 1.8.1

• add2csv(): create a new entry in the documentation different than clearNN (minor bug)

NNbenchmark 1.8

• NNdatasets now includes for each dataset the corresponding fmlaNN and nparNN
• prepareZZ(Zsm = ...) adds has a new item to the final list:
• Zxy is a scaled or not scaled data.frame or matrix, depending the zmd and scale parameters

NNbenchmark 1.7.2

• DESCRIPTION Imports: pkgload
• clearNN() => uses pkgload::unload() to unload poperly S4 classes and dll

NNbenchmark 1.7.1

• clearNN() => "tools:rstudio" added to the default packages (NULL) not be detached

NNbenchmark 1.7

• add2csv()
• funRMSE(), funMAE()
• MAE (Mean absolute Error) added to timer_class and data.frame
• clearNN() => tryCatch(...) added

NNbenchmark 1.6

• clearNN()

NNbenchmark 1.5

• ht(..., l = 3)
• ht9()
• NNdatasets
• prepareZZ(..., scale =) replaces prepareZZ(..., normalize) as 'scale' is the usual word
• for (mean, standard deviation) adjustment in R.
• prepareZZ() now returns the formula y ~ x or y ~ x1 + ... + xn for any number of columns

NNbenchmark 1.4.1

• Function plotNN(, TF) + lipoNN(, TF) replaces if (TRUE/FALSE) plot
• Examples prepareZZ() rewritten accordingly

NNbenchmark 1.4

• Function plotNN() + lipoNN()
• Examples prepareZZ() rewritten accordingly

NNbenchmark 1.3

• Function prepareZZ()
• timeR: add a column RMSE

NNbenchmark 1.2

• Function prepareZ()
• DESCRIPTION Suggests: brnn, validann

NNbenchmark 1.1

• Functions and classes timeR

NNbenchmark 1.0

• Functions cc() and ht()
• Datasets are now uNNdatasets.rda and mNNdatasets.rda 'dataNNbenchmark'
• LazyData: true' in DESCRIPTION file