Package 'NNbenchmark'

Description

Datasets and functions to benchmark (convergence, speed, ease of use) R packages dedicated to regression with neural networks (no classification in this version). The templates for the tested packages are available at <https://github.com/pkR-pkR/NNbenchmarkTemplates> and <https://theairbend3r.github.io/NNbenchmarkWeb/index.html>. The submitted article to the R-Journal can be read at <https://www.inmodelia.com/gsoc2020.html>.

Examples

ds <- grep("^[m,u]", ls("package:NNbenchmark"), value = TRUE); ds
t(sapply(ds, function(x) dim(get(x)))) 
NNdataSummary(NNdatasets)
plot(uGauss2)
pairs(mIshigami)
ht(NNdatasets, n = 2, l = 6)

Description

Create or append a data.frame to a csv file. Column names are added at creation and ignored at the append steps.

Usage

add2csv(x, file = "results.csv", dir = ".")

Arguments

Value

Nothing in the console. A csv file on the disk.

Examples

results_csv <- tempfile("results", fileext = ".csv")
x <- data.frame(a = 1:3, b = 4:6) 
add2csv(x, file = results_csv) 
add2csv(x*10, file = results_csv) 
add2csv(x*100, file = results_csv) 
read.csv(file = results_csv)

Description

A multivariate dataset (x1, x2, x3, x4, x5, x6, y) of class data.frame and dim 20000 x 7 to be fitted by a neural network with 5 hidden neurons (41 parameters).

References

Inspired by page 29 of Wood, S. N. (2011). Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 73(1), 3-36. https://people.bath.ac.uk/man54/SAMBa/ITTs/ITT2/EDF/REMLWood2009.pdf

Examples

ht(bWoodN1)

Description

An intermediate function between c() and list(). Combine all terms in one single list. The result can be used by do.call().

Usage

cc(char, ...)

Arguments

Value

A list with the objects concatenated.

Examples

lst <- list(yaxt = "s", side = 2, col = 1:3) ; lst
dfr <- data.frame(x = 5:9, y = 10:14) ; dfr

## With c(), the list is returned at its given position
c(lst, at = 7, labels = c("0", "0.5", "1"), dfr = dfr)
c(at = 7, labels = c("0", "0.5", "1"), lst, dfr = dfr)

## With cc(), the unnamed list is always returned in first position
cc(lst, at = 7, labels = c("0", "0.5", "1"), dfr = dfr)
cc(at = 7, dfr = dfr, labels = c("0", "0.5", "1"), lst)

## Some similarities and differences between c() and cc()
c( 1:5, y = 2:6, col = 2, lwd = 2)
cc(1:5, y = 2:6, col = 2, lwd = 2)

c( x = 1:5, y = 2:6, col = 2, lwd = 2)
cc(x = 1:5, y = 2:6, col = 2, lwd = 2)

## Regular function and do.call()
plot(x = 1:5, y = 2:6, col = 2, lwd = 2)
do.call( plot , cc(x = 1:5, y = 2:6, col = 3, lwd = 3, cex = 3))
do.call("plot", cc(x = 1:5, y = 2:6, col = 4, lwd = 4, cex = 4))

Description

clearNN detachs ZZ and the packages loaded for the evaluation.

detachNN detachs ZZ.

Usage

clearNN(donotremove, donotdetach = NULL)

detachNN()

Arguments

Description

This is a modified version of the timeR package for an internal use. Full credit is to Yifu Yan, the author of the timeR package.

createTimer creates a timer object.

getTimer returns a data frame with all records saved by the timer object. Columns in the data.frame are: event, start, end, duration, RMSE, MAE, stars, params, comment.

Usage

createTimer(verbose = TRUE)

getTimer(object)

Arguments

Value

An (invisible) object of R6 class for createTimer. A data.frame for getTimer.

Examples

## Create a timer object. Record events. Get all records.
timeTT <- createTimer(FALSE) # print is disabled
timeTT <- createTimer()      # print is enabled
timeTT$start("event1")
Sys.sleep(1)
timeTT$stop("event1", RMSE = 1, MAE = 1.3, stars = "*", 
            params = "maxiter=100, lr=0.01", comment = "OK for 1",  
            printmsg = TRUE)
timeTT$start("event2")
Sys.sleep(2)
timeTT$stop("event2", RMSE = 2, MAE = 2.6, stars = "**",  
            params = "maxiter=1000, lr=0.001", comment = "OK for 2",  
            printmsg = FALSE)
getTimer(timeTT)

Description

Calculate the Root Mean Squared Error (RMSE), the Mean Squared Error (MSE), the Mean Absoluter Error (MAE), and the Worst Absolute Error (WAE). The result is rounded to 4 digits by default. Apply na.rm = TRUE

Usage

funRMSE(y_pred, y0, dgts = 4)

funMSE(y_pred, y0, dgts = 4)

funMAE(y_pred, y0, dgts = 4)

funWAE(y_pred, y0, dgts = 4)

Arguments

Value

A numeric value, either the RMSE, MSE, MAE, or WAE.

Examples

y0 <- 1:19
y_pred <- y0 + rnorm(length(y0), sd = 0.3)
funRMSE(y_pred, y0)
funMSE( y_pred, y0)
funMAE( y_pred, y0)
funWAE( y_pred, y0)

Description

Concatenates head(n) and tail(n) rows and subset with m columns. Works also with array, list and matrix in data.frame. Keeps the data.table format (and add a timezone by default). See the matsindf package for matrix(ces) in tibble.

Usage

ht(x, n = 3, m = 4, p = 2, l = 2, names = TRUE, LTT = c("x", "L",
  "N", "M", "P", "Q", "R", "S", "T"))

ht9(x, n = 3, m = 9999, p = 2, l = 2, names = TRUE, LTT = c("x",
  "L", "N", "M", "P", "Q", "R", "S", "T"))

Arguments

Value

An object of the same class than x but much shorter.

Examples

### Vector, data.frame, array
ht(1:100, names = FALSE)
ht(1:100, LTT = "z")

ht9(mtcars); dim(mtcars)

arr4 <- array(1:1680, c(8,7,6,5))
ht(arr4, n=1, p=1, names = FALSE)
ht(arr4, n=1, p=1, names = TRUE, LTT = c("x","L","X","Y","Z","T"))

### List of matrices
lstmat <- rep(list(matrix(1:100, 10)), 8)
for (i in seq_along(lstmat)) lstmat[[i]] <- lstmat[[i]] *i
lstmat
ht(lstmat, n = 2, m = 2, l = 2, names = FALSE)
ht(lstmat, n = 2, m = 3, l = 1, names = TRUE)
ht(lstmat, n = 2, m = 3, l = 1, LTT = c("x","L","X","Y","Z"))

### Data.frame with matrices inside.
### See For instance data("gasoline", package = "pls")

## Colnames on matrix B but not on matrix C. Protected data.frame.
B <- matrix(101:160, 10); colnames(B) <- paste0(1:6, "b"); B
C <- matrix(101:160, 10); C
dfrmat <- data.frame(A = 1:10, B = I(B), C = I(C), D = 11:20); dfrmat
colnames(dfrmat)

## Matrix columns are controlled by m.
## Unnamed C matrix columns have old values but new names. Be aware!
ht(dfrmat, n = 2, m = 1, l = 2, names = FALSE) # Original C.6 is now C.2
ht(dfrmat, n = 2, m = 1, l = 2, names = TRUE)  # Names keep original ranks

## Data.frame columns are controlled by l.
ht(dfrmat, n = 3, m = 2, l = 1, names = TRUE)

Description

A multivariate dataset (x1, x2, x3, y) of class matrix and dim 500 x 4 to be fitted by a neural network with 5 hidden neurons (26 parameters).

References

Dette, H., & Pepelyshev, A. (2010). Generalized Latin hypercube design for computer experiments. Technometrics, 52(4).

See also https://www.sfu.ca/~ssurjano/detpep10curv.html

Examples

ht(mDette)
pairs(mDette)

Description

A multivariate dataset (x1, x2, x3, x4, x5, y) of class matrix and dim 500 x 6 to be fitted by a neural network with 5 hidden neurons (36 parameters).

References

Friedman, J. H., Grosse, E., & Stuetzle, W. (1983). Multidimensional additive spline approximation. SIAM Journal on Scientific and Statistical Computing, 4(2), 291-301.

See also https://www.sfu.ca/~ssurjano/fried.html

Examples

ht(mFriedman)
pairs(mFriedman)

Description

A multivariate dataset (x1, x2, x3, y) of class matrix and dim 500 x 4 to be fitted by a neural network with 10 hidden neurons (51 parameters).

References

Ishigami, T., & Homma, T. (1990, December). An importance quantification technique in uncertainty analysis for computer models. In Uncertainty Modeling and Analysis, 1990. Proceedings., First International Symposium on (pp. 398-403). IEEE.

See also https://www.sfu.ca/~ssurjano/ishigami.html

Examples

ht(mIshigami)
pairs(mIshigami)

Description

A multivariate dataset (x1, x2, x3, x4, x5, y) of class matrix and dim 153 x 6 to be fitted by a neural network with 3 hidden neurons (22 parameters). This dataset was used to teach neural networks at ESPCI from 2003 to 2013 and is available in the software Neuro One.

References

Neuro One https://www.inmodelia.com/software.html

Examples

ht(mRef153)
pairs(mRef153)

Description

NNbigdatasets is a list with the big datasets presented in this package and the recommended number of hidden neurons for each neural network model.

• bWoodN1: 5 neurons.

Each item of the list is itself a list with 5 components:

• ds: character. The name of the dataset.

• neur: integer. The recommanded number of hidden neurons in the NN model and in fmlaNN.

• nparNN: integer. The number of parameters in fmlaNN.

• fmlaNN: the formula of the corresponding neural network, with tanh() as the activation function in the hidden layer.

• Z: matrix or data.frame. The dataset itself.

Using attach() and detach() gives a direct access to these items.

Examples

ht(NNbigdatasets)

NNdataSummary(NNbigdatasets)

Description

NNdatasets is a list with the 12 datasets presented in this package and the recommended number of hidden neurons for each neural network model.

• mDette: 5 neurons.

• mFriedman: 5 neurons.

• mIshigami: 10 neurons.

• mRef153: 3 neurons.

• uDmod1: 6 neurons.

• uDmod2: 5 neurons.

• uDreyfus1: 3 neurons.

• uDreyfus2: 3 neurons.

• uGauss1: 5 neurons.

• uGauss2: 4 neurons.

• uGauss3: 4 neurons.

• uNeuroOne: 2 neurons.

Each item of the list is itself a list with 5 components:

• ds: character. The name of the dataset.

• neur: integer. The recommanded number of hidden neurons in the NN model and in fmlaNN.

• nparNN: integer. The number of parameters in fmlaNN.

• fmlaNN: the formula of the corresponding neural network, with tanh() as the activation function in the hidden layer.

• Z: matrix or data.frame. The dataset itself.

Using attach() and detach() gives a direct access to these items.

Examples

ht(NNdatasets, n = 2, l = 6)

NNdataSummary(NNdatasets)

Description

NNdataSummary summarizes the information of the 12 datasets listed in NNdatasets.

Usage

NNdataSummary(NNdatasets)

Arguments

Value

A data.frame with 12 rows and 5 columns: (dataset) name, n_rows, n_inputs, n_neurons, n_parameters.

Examples

NNdataSummary(NNdatasets)

Description

Summarize measures of fit and time for a single training. Measures of fit include the Root Mean Squared Error (RMSE), the Mean Squared Error (MSE), the Mean Absolute Error (MAE), and the Worst Absolute Error (WAE) rounded by default to 4 digits and set to na.rm = TRUE. See more at funRMSE. The summary can also include the results of time from getTimer in NNbenchmark or the result of timediff.

Usage

NNsummary(y_pred, y0, time, dgts = 4)

Arguments

Value

A vector of RMSE, MSE, MAE, WAE, and time values for a single iteration.

Examples

## With 2019 legacy code, no longer usable with 2020 trainPredict
old <- options("digits.secs" = 4)
timeTT <- createTimer()

timeTT$start("event")
y0 <- 1:19
y_pred <- y0 + rnorm(length(y0), sd = 0.3)
timeTT$stop("event") 

time <- getTimer(timeTT)

NNsummary(y_pred, y0, time[,4], 4)

## With 2020 code
timestart()
y0 <- 1:19
y_pred <- y0 + rnorm(length(y0), sd = 0.3)
time <- timediff()

NNsummary(y_pred, y0, time, 4)

options(old)

Description

An implementation with do.call so that any neural network function that fits the format can be tested.

In trainPredict_1mth1data, a neural network is trained on one dataset and then used for predictions, with several functionalities. Then, the performance of the neural network is summarized.

trainPredict_1data serves as a wrapper function for trainPredict_1mth1data for multiple methods.

trainPredict_1pkg serves as a wrapper function for trainPredict_1mth1data for multiple datasets.

Usage

trainPredict_1mth1data(dset, method, trainFUN, hyperparamFUN, predictFUN,
  summaryFUN, prepareZZ.arg = list(), nrep = 5, doplot = FALSE,
  plot.arg = list(col1 = 1:nrep, lwd1 = 1, col2 = 4, lwd2 = 3), pkgname,
  pkgfun, csvfile = FALSE, rdafile = FALSE, odir = ".", echo = FALSE,
  echoreport = FALSE, appendcsv = TRUE, ...)

trainPredict_1data(dset, methodlist, trainFUN, hyperparamFUN, predictFUN,
  summaryFUN, closeFUN, startNN = NA, prepareZZ.arg = list(), nrep = 5,
  doplot = FALSE, plot.arg = list(), pkgname = "pkg", pkgfun = "train",
  csvfile = FALSE, rdafile = FALSE, odir = ".", echo = FALSE,
  lib.loc = NULL, ...)

trainPredict_1pkg(dsetnum, pkgname = "pkg", pkgfun = "train", methodvect,
  prepareZZ.arg = list(), summaryFUN, nrep = 5, doplot = FALSE,
  plot.arg = list(), csvfile = FALSE, rdafile = FALSE, odir = ".",
  echo = FALSE, appendcsv = TRUE, parallel = "no", ncpus = 1,
  cl = NULL, lib.loc = NULL, ...)

Arguments

Value

An array with values as in NNsummary including each repetition, with options for plots and output files

Examples

nrep <- 2       
odir <- tempdir()

### Package with one method/optimization algorithm
library("brnn")
brnn.method <- "gaussNewton"
hyperParams.brnn <- function(optim_method, ...) {
  return(list(iter = 200))
  }
brnn.prepareZZ <- list(xdmv = "m", ydmv = "v", zdm = "d", scale = TRUE)

NNtrain.brnn   <- function(x, y, dataxy, formula, neur, optim_method, hyperParams,...) {
  hyper_params <- do.call(hyperParams.brnn, list(brnn.method))
  iter  <- hyper_params$iter
  
  NNreg <- brnn::brnn(x, y, neur, normalize = FALSE, epochs = iter, verbose = FALSE)
  return(NNreg)
  }
NNpredict.brnn <- function(object, x, ...) { predict(object, x) }
NNclose.brnn <- function(){
  if("package:brnn" %in% search())
    detach("package:brnn", unload=TRUE)
  }


res <- trainPredict_1pkg(1:2, pkgname = "brnn", pkgfun = "brnn", brnn.method,
                         prepareZZ.arg = brnn.prepareZZ, nrep = nrep, doplot = TRUE,
                         csvfile = FALSE, rdafile = FALSE, odir = odir, echo = FALSE)
                         
### Package with more than one method/optimization algorithm
library(validann)
validann.method <- c("Nelder-Mead", "BFGS", "CG", "L-BFGS-B", "SANN")
hyperParams.validann <- function(optim_method, ...) {
  if(optim_method == "Nelder-Mead")  { maxiter <- 10000 } 
  if(optim_method == "BFGS")         { maxiter <- 200   }
  if(optim_method == "CG")           { maxiter <- 1000  }
  if(optim_method == "L-BFGS-B")     { maxiter <- 200   }
  if(optim_method == "SANN")         { maxiter <- 1000  }
  return(list(iter = maxiter, method = optim_method, params))
  }
validann.prepareZZ <- list(xdmv = "m", ydmv = "m", zdm = "d", scale = TRUE)

NNtrain.validann <- function(x, y, dataxy, formula, neur, optim_method, hyperParams, ...) {
  hyper_params <- do.call(hyperParams, list(optim_method, ...))
  iter <- hyper_params$iter
  method <- hyper_params$method
  
  NNreg <- validann::ann(x, y, size = neur, method = method, maxit = iter)
  return (NNreg)
  }
NNpredict.validann <- function(object, x, ...) { predict(object, x) }
NNclose.validann <- function() {
  if("package:validann" %in% search())
  detach("package:validann", unload=TRUE)
  }

res <- trainPredict_1pkg(1:2, pkgname = "validann", pkgfun = "ann", validann.method,
                         repareZZ.arg = validann.prepareZZ, nrep = nrep, doplot = FALSE,
                         csvfile = TRUE, rdafile = TRUE, odir = odir, echo = FALSE)

Description

plotNN uses the parameter uni to launch the plot() function either for a univariate dataset (x, y_pred) or for a multivariate dataset (y, y_pred).

lipoNN uses the parameter uni to launch either lines() for an univariate dataset or points() for a multivariate dataset.

See the examples in prepareZZ.

Usage

plotNN(xory, y0, uni, TF = TRUE, ...)

lipoNN(xory, y_pred, uni, TF = TRUE, ...)

Arguments

Value

NULL in the console. An initial plot or some added lines/points.

Description

This function modifies a dataset to the format required by a training function: data.frame, matrix or vector (numeric), pre-normalization.

Usage

prepareZZ(Z, xdmv = "m", ydmv = "v", zdm = "d", scale = FALSE)

Arguments

Value

The output is a list, usually named ZZ, with the following items:

• Zxy: the original or scaled Z in the desired format (data.frame, matrix).

• x: the original or scaled x in the desired format (data.frame, matrix, vector).

• y: the original or scaled y in the desired format (data.frame, matrix, vector).

• xory: the original x or y (as vector).

• y0: the original y (as vector).

• xm0: the mean(s) of the original x.

• ym0: the mean of the original y.

• xsd0: the standard deviation(s) of the original x.

• ysd0: the standard deviation of the original y.

• uni: the univariate (TRUE) or multivariate (FALSE) status of x (Z).

• fmla: the formula y ~ x or y ~ x1 + x2 + .. + xn where n is the number of inputs variables.

The use of attach() and detach() gives direct access to the modified values of ZZ. See the examples.

Examples

library("brnn")
library("validann")

maxit <- 200  # increase this number to get more accurate results with validann:ann
TF    <- TRUE # display the plots

### UNIVARIATE DATASET
Z     <- uGauss2
neur  <- 4

## brnn
ZZ <- prepareZZ(Z, xdmv = "m", ydmv= "v", scale = FALSE) ; ht(ZZ) 
attach(ZZ)
y_pred <- ym0 + ysd0*predict(brnn(x, y, neur))
plotNN(xory, y0, uni, TF)
lipoNN(xory, y_pred, uni, TF, lwd = 4, col = 2)
ym0 ; ysd0
detach(ZZ) ; rm(y_pred)


## validann
ZZ <- prepareZZ(Z, xdmv = "m", ydmv= "v", scale = TRUE) ; ht(ZZ)
attach(ZZ)
y_pred <- ym0 + ysd0*predict(validann::ann(x, y, neur, maxit = maxit))
lipoNN(xory, y_pred, uni, TF, lwd = 4, col = 3)
ym0 ; ysd0
detach(ZZ) ; rm(y_pred)


### UNIVARIATE DATASET + LOOP
nruns  <- 10

## brnn
ZZ <- prepareZZ(Z, xdmv = "m", ydmv= "v", scale = FALSE) ; ht(ZZ) 
attach(ZZ)
Zreg <- list() ; Zreg
for (i in 1:nruns) Zreg[[i]] <- brnn::brnn(x, y, neur) 
m      <- matrix(sapply(Zreg, function(x) x$Ed) , ncol=1) ; m
best   <- which(min(m) == m)[1] ; best
y_pred <- ym0 + ysd0*predict(Zreg[[best]])
plotNN(xory, y0, uni, TF)
lipoNN(xory, y_pred, uni, TF, lwd = 4, col = 2)
detach(ZZ) ; rm(y_pred)

## validann
ZZ <- prepareZZ(Z, xdmv = "m", ydmv= "v", scale = TRUE) ; ht(ZZ)
attach(ZZ)
Zreg <- list()
for (i in 1:nruns) Zreg[[i]] <- validann::ann(x, y, size = neur, maxit = maxit) 
m      <- matrix(sapply(Zreg, function(x) x$value), ncol=1) ; m
best   <- which(min(m) == m)[1] ; best
y_pred <- ym0 + ysd0*predict(Zreg[[best]])
lipoNN(xory, y_pred, uni, TF, lwd = 4, col = 4)
detach(ZZ) ; rm(y_pred)


### MULTIVARIATE DATASET
Z    <- mDette
neur <- 5

## brnn
ZZ <- prepareZZ(Z, xdmv = "m", ydmv= "v", scale = FALSE) ; ht(ZZ)
attach(ZZ)
y_pred <- ym0 + ysd0*predict(brnn::brnn(x, y, neur))
plotNN(xory, y0, uni, TF)
lipoNN(xory, y_pred, uni, TF, lwd = 4, col = 2)
ym0 ; ysd0
detach(ZZ) ; rm(y_pred)

## validann
ZZ <- prepareZZ(Z, xdmv = "m", ydmv= "v", scale = TRUE) ; ht(ZZ)
attach(ZZ)
y_pred <- ym0 + ysd0*predict(validann::ann(x, y, neur, maxit = maxit))
lipoNN(xory, y_pred, uni, TF, lwd = 4, col = 3)
ym0 ; ysd0
detach(ZZ) ; rm(y_pred)


### INSIDE A FUNCTION
plotds <- function(Z, xdmv = "m", ydmv = "v", scale = FALSE, neurons = 3, col = 2) {
    ZZ <- prepareZZ(Z, xdmv = xdmv, ydmv= ydmv, scale = scale) 
    attach(ZZ) ; on.exit(detach(ZZ))
    y_pred <- ym0 + ysd0*predict(brnn::brnn(x, y, neurons))
    plotNN(xory, y0, uni, TF)
    lipoNN(xory, y_pred, uni, TF, lwd = 4, col = col)
    print(ht(x))
    print(ht(y))
}
plotds(uNeuroOne, scale = FALSE, neurons = 2, col = 2)
plotds(uNeuroOne, scale = TRUE,  neurons = 3, col = 3)

plotds(mFriedman, scale = TRUE,  neurons = 5, col = 4)

Description

timer is a R6 Class that represent a timer. This is a modified version of the timeR package for an internal use. Full credit is to Yifu Yan, the author of the timeR package.

Value

getTimer returns a data frame with all records saved by the timer object. Columns in the data.frame are: event, start, end, duration, RMSE, MAE, stars, params, comment.

Public Methods

initialize(time,event,verbose,eventTable)

Initialize a timer object. You can also use createTimer() function to initialize a timer object.

start(eventName)

Start timing for a event, eventName should be a string

stop(eventName)

Stop timing for a event.

getTimer()

Get/Print a data.frame with all records.

removeEvent(eventName)

Remove an given row in the eventTable.

toggleVerbose()

Toggle between TRUE and FALSE for verbose

getStartTime()

Get start time for a selected event.

getStopTime()

Get stop time for a selected event.

getDuration()

Get duration for a selected event.

getRMSE()

Get the RMSE for a selected event.

getMAE()

Get the MAE for a selected event.

getStars()

Get stars for a selected event.

getParams()

Get params for a selected event.

getComment()

Get comment for a selected event.

getEventf()

Get entire row for a selected event.

print()

Custom print method for timer class. However, you don't need to use this function to generate custom printing. Custom printing is triggered by default.

Private Methods

slprint(msg, flag = self$verbose)

A function that controls whether to print extra message.

Public fields

Active bindings

Methods

Public methods

• timeR$new()

• timeR$start()

• timeR$stop()

• timeR$getTimer()

• timeR$removeEvent()

• timeR$toggleVerbose()

• timeR$getStartTime()

• timeR$getStopTime()

• timeR$getDuration()

• timeR$getRMSE()

• timeR$getMAE()

• timeR$getStars()

• timeR$getParams()

• timeR$getComment()

• timeR$getEvent()

• timeR$print()

• timeR$clone()

Method new()

Usage

timeR$new(verbose = TRUE)

Method start()

Usage

timeR$start(eventName)

Method stop()

Usage

timeR$stop(eventName, RMSE = NA_real_, MAE = NA_real_,
  stars = NA_character_, params = NA_character_, comment = NA_character_,
  printmsg = TRUE)

Method getTimer()

Usage

timeR$getTimer(...)

Method removeEvent()

Usage

timeR$removeEvent(eventName)

Method toggleVerbose()

Usage

timeR$toggleVerbose(...)

Method getStartTime()

Usage

timeR$getStartTime(eventName)

Method getStopTime()

Usage

timeR$getStopTime(eventName)

Method getDuration()

Usage

timeR$getDuration(eventName)

Method getRMSE()

Usage

timeR$getRMSE(eventName)

Method getMAE()

Usage

timeR$getMAE(eventName)

Method getStars()

Usage

timeR$getStars(eventName)

Method getParams()

Usage

timeR$getParams(eventName)

Method getComment()

Usage

timeR$getComment(eventName)

Method getEvent()

Usage

timeR$getEvent(eventName)

Method print()

Usage

timeR$print(...)

Method clone()

The objects of this class are cloneable with this method.

Usage

timeR$clone(deep = FALSE)

Arguments

Examples

timer <- createTimer()
timer$start("event1")
## put some codes in between, for instance
Sys.sleep(1)
timer$stop("event1", RMSE = 1, MAE = 1.3, stars = "*", 
           params = "maxiter=100, lr=0.01", comment = "OK for 1",  
           printmsg = TRUE)

timer$start("event2")
## put some codes in between, for instance
Sys.sleep(2)
timer$stop("event2", RMSE = 2, MAE = 2.6, stars = "**",  
           params = "maxiter=1000, lr=0.001", comment = "OK for 2",  
           printmsg = FALSE)

table1 <- getTimer(timer)
timer$toggleVerbose() # set verbose to FALSE as default is TRUE
table1 # print all records in a data frame

## get attributes for selected events
timer$getStartTime("event1")
timer$getStopTime("event1")
timer$getDuration("event1")
timer$getComment("event1")
timer$getEvent("event1")

Description

timestart starts the timer and saved the value in an object named time0 stored in .GlobalEnv.

timediff stops the timer, remove the time0 objet from .GlobalEnv and prints the duration in seconds between the two events.

timestart and timediff are fully independant from the R6 class timeR and the objects createTimer or getTimer. They use proc.time instead.

Usage

timestart()

timediff()

Value

A single numeric value that represents a duration in seconds.

Examples

timestart()
Sys.sleep(2)
timediff()

Description

An univariate dataset (x, y) of class data.frame and dim 51 x 2 to be fitted by a neural network with 6 hidden neurons (19 parameters). The parameters are highly correlated and singular Jacobian matrices often appear. A difficult dataset.

Examples

ht(uDmod1)
plot(uDmod1)

Description

An univariate dataset (x, y) of class data.frame and dim 51 x 2 to be fitted by a neural network with 5 hidden neurons (16 parameters).

Examples

ht(uDmod2)
plot(uDmod2)

Description

An univariate dataset (x, y) of class data.frame and dim 51 x 2 to be fitted by a neural network with 3 hidden neurons (10 parameters). This dataset was used to teach neural networks at ESPCI from 1991 to 2013. It usually appeared in the very first slides. This is a combination of 3 pure tanh() functions without noise. The Jacobian matrix is singular at the target parameter values and many algorithms could fail.

References

Dreyfus, G., ESPCI https://www.neurones.espci.fr

Examples

ht(uDreyfus1)
plot(uDreyfus1)

Description

An univariate dataset (x, y) of class data.frame and dim 51 x 2 to be fitted by a neural network with 3 hidden neurons (10 parameters). This dataset was used to teach neural networks at ESPCI from 1991 to 2013. It usually appeared in the very first slides. This is a combination of 3 pure tanh() functions with a small noise. Due to the noise, the Jacobian matrix is not singular at the target parameter values. All algorithms should find the target parameter values.

References

Dreyfus, G., ESPCI https://www.neurones.espci.fr

Examples

ht(uDreyfus2)
plot(uDreyfus2)

Description

An univariate dataset (x, y) of class data.frame and dim 250 x 2 to be fitted by a neural network with 5 hidden neurons (16 parameters).

References

Rust, B., NIST (1996) https://www.inmodelia.com/gsoc2020-redirection.html with redirection to the (slow) NIST page.

Examples

ht(uGauss1)
plot(uGauss1)

Description

An univariate dataset (x, y) of class data.frame and dim 250 x 2 to be fitted by a neural network with 4 hidden neurons (13 parameters).

References

Rust, B., NIST (1996) https://www.inmodelia.com/gsoc2020-redirection.html with redirection to the (slow) NIST page.

Examples

ht(uGauss2)
plot(uGauss2)

Description

An univariate dataset (x, y) of class data.frame and dim 250 x 2 to be fitted by a neural network with 4 hidden neurons (13 parameters).

References

Rust, B., NIST (1996) https://www.inmodelia.com/gsoc2020-redirection.html with redirection to the (slow) NIST page.

Examples

ht(uGauss3)
plot(uGauss3)

Description

An univariate dataset (x, y) of class data.frame and dim 51 x 2 to be fitted by a neural network with 2 hidden neurons (7 parameters). This dataset was used to teach neural networks at ESPCI from 1991 to 2013 and is available in the software Neuro One.

References

Dreyfus, G., ESPCI https://www.neurones.espci.fr

Neuro One https://www.inmodelia.com/software.html

Examples

ht(uNeuroOne)
plot(uNeuroOne)