mlpWeightDecayML - learnFuncParams decay parameter at wrong index

[smlek]

when using mlpWeightDecayML and mlpWeightDecay methods and specifying decay parameter in tuneGrid, the decay value is passed incorrectly during the call to RSNNS::mlp() learnFuncParams.

Decay value should be at learnFuncParams[2], but instead it shows up in learnFuncParams[3].
This is an error in mlpWeightDecayML$fit

mlpWeightDecayML$fit
.
.
if (any(names(theDots) == "learnFuncParams")) 
        prms <- theDots$learnFuncParams

        prms[3] <- param$decay

This should be prms[2] <- param$decay

debugging call to mlp() gives:
learnFunc = "BackpropWeightDecay", learnFuncParams = c(0.01, 5e-05, 1e-05, 0), size = 4, ...

Minimal dataset:

library(caret)
data("infert")

Minimal, runnable code:

cvCtrl <- trainControl(method = "none", number = 1)
tune_grid <- expand.grid(layer1 = c(4),
                         layer2 = 0,
                         layer3 = 0,
                         decay = 0.00001
                         )

mlpwdMLModel <- train(x = infert[, c(2, 3, 4, 6)],
                 y = infert[, 5],
                 #y = factor(trainOutputs, levels=c("0","1"), labels=c("Down","Up")),
                 method = "mlpWeightDecayML",
                 learnFuncParams=c(0.01, .00005, 0, 0),
                 #tuneLength = 30,
                 #metric = "Accuracy",
                 trControl = cvCtrl,
                 tuneGrid = tune_grid,
                 maxit=500,
                 linOut = FALSE            
                 )
)
print(mlpwdMLModel)

Session Info:

>sessionInfo()

R version 3.2.3 (2015-12-10)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 7 x64 (build 7601) Service Pack 1

locale:
[1] LC_COLLATE=English_United States.1252 LC_CTYPE=English_United States.1252
[3] LC_MONETARY=English_United States.1252 LC_NUMERIC=C
[5] LC_TIME=English_United States.1252

attached base packages:
[1] stats graphics grDevices utils datasets methods base

other attached packages:
[1] caret_6.0-64 ggplot2_2.0.0 lattice_0.20-33 RSNNS_0.4-7 Rcpp_0.12.2

loaded via a namespace (and not attached):
[1] knitr_1.12 magrittr_1.5 splines_3.2.3 MASS_7.3-45 munsell_0.4.3
[6] colorspace_1.2-6 foreach_1.4.3 minqa_1.2.4 stringr_1.0.0 car_2.1-1
[11] plyr_1.8.3 tools_3.2.3 parallel_3.2.3 nnet_7.3-11 pbkrtest_0.4-6
[16] grid_3.2.3 gtable_0.1.2 nlme_3.1-124 mgcv_1.8-10 quantreg_5.21
[21] MatrixModels_0.4-1 iterators_1.0.8 lme4_1.1-11 Matrix_1.2-3 nloptr_1.0.4
[26] reshape2_1.4.1 codetools_0.2-14 stringi_1.0-1 scales_0.3.0 stats4_3.2.3
[31] SparseM_1.7

[topepo]

True. I've committed the fixes and they will be in the next version.

Thanks,

Max