1 Setup

1.1 Packages and options

library(NNbenchmark)
library(kableExtra)
options(scipen = 999)

1.2 Datasets to test

NNdataSummary(NNdatasets)

##      Dataset n_rows n_inputs n_neurons n_parameters
## 1     mDette    500        3         5           26
## 2  mFriedman    500        5         5           36
## 3  mIshigami    500        3        10           51
## 4    mRef153    153        5         3           22
## 5     uDmod1     51        1         6           19
## 6     uDmod2     51        1         5           16
## 7  uDreyfus1     51        1         3           10
## 8  uDreyfus2     51        1         3           10
## 9    uGauss1    250        1         5           16
## 10   uGauss2    250        1         4           13
## 11   uGauss3    250        1         4           13
## 12 uNeuroOne     51        1         2            7

1.3 RSNNS trainPredict arguments - inputs x, y

if(dir.exists("D:/GSoC2020/Results/2020run03/"))
{  
  odir <- "D:/GSoC2020/Results/2020run03/"
}else if(dir.exists("~/Documents/recherche-enseignement/code/R/NNbenchmark-project/NNtempresult/"))
{  
  odir <- "~/Documents/recherche-enseignement/code/R/NNbenchmark-project/NNtempresult/"
}else
  odir <- "~"

nrep <- 10
maxit2ndorder  <-    200
maxit1storderA <-   1000
maxit1storderB <-  10000
maxit1storderC <- 100000

#library(RSNNS)
RSNNS.method <- c("Rprop","BackpropBatch","BackpropChunk","BackpropMomentum",
                  "BackpropWeightDecay","Quickprop","SCG","Std_Backpropagation")
hyperParams.RSNNS <- function(optim_method, ...) {
    
    if (optim_method == "Std_Backpropagation")   {iter <- maxit1storderA} 
    if (optim_method == "BackpropBatch")         {iter <- maxit1storderB}
    if (optim_method == "BackpropChunk")         {iter <- maxit1storderA}
    if (optim_method == "BackpropMomentum")      {iter <- maxit1storderA}
    if (optim_method == "BackpropWeightDecay")   {iter <- maxit1storderA}
    if (optim_method == "Rprop")                 {iter <- maxit1storderA}
    if (optim_method == "Quickprop")             {iter <- maxit1storderB}
    if (optim_method == "SCG")                   {iter <- maxit1storderA}

    out <- list(iter = iter)
    return(out)
}
NNtrain.RSNNS <- function(x, y, dataxy, formula, hidden_neur, optim_method, hyperParams, NNfullformula, NNparam,...) {
    
    hyper_params <- do.call(hyperParams, list(optim_method, ...))
    if (optim_method == "Std_Backpropagation" || optim_method == "BackpropBatch"){
          NNreg <- RSNNS::mlp(x, y, size = hidden_neur, learnFunc = optim_method, learnFuncParams = c(0.1), #lr 
                        maxit = hyper_params$iter, linOut = TRUE)
    } else {
          NNreg <- RSNNS::mlp(x, y, size = hidden_neur, learnFunc = optim_method,
                        maxit = hyper_params$iter, linOut = TRUE)
    }
    
    return (NNreg)
}
NNpredict.RSNNS <- function(object, x, ...)
  predict(object, x)
NNclose.RSNNS <- function()
  if("package:RSNNS" %in% search())
    detach("package:RSNNS", unload=TRUE)
RSNNS.prepareZZ <- list(xdmv = "m", ydmv = "m", zdm = "d", scale = TRUE)


if(FALSE)
res <- train_and_predict_1data(1, RSNNS.method, "NNtrain.RSNNS", "hyperParams.RSNNS", "NNpredict.RSNNS", 
                               NNsummary, "NNclose.RSNNS", NA, RSNNS.prepareZZ, nrep=5, echo=TRUE, doplot=FALSE,
                               pkgname="RSNNS", pkgfun="mlp", rdafile=TRUE, odir=odir)

2 Launch package’s trainPredict

res <- trainPredict_1pkg(1:12, pkgname = "RSNNS", pkgfun = "mlp", RSNNS.method,
  prepareZZ.arg = RSNNS.prepareZZ, nrep = nrep, doplot = FALSE,
  csvfile = TRUE, rdafile = TRUE, odir = odir, echo = FALSE)

3 Results

#print(res)
kable(apply(res[,,1,], c(3,1), min), caption = "Rprop")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

Rprop
	RMSE	MSE	MAE	WAE	time
mDette	0.5709	0.3259	0.4394	2.1676	0.58
mFriedman	0.0263	0.0007	0.0207	0.0853	0.58
mIshigami	0.8069	0.6510	0.5909	3.3266	0.71
mRef153	3.2667	10.6712	2.3033	13.4889	0.16
uDmod1	0.0578	0.0033	0.0445	0.1814	0.08
uDmod2	0.0744	0.0055	0.0623	0.1463	0.07
uDreyfus1	0.0617	0.0038	0.0450	0.1563	0.06
uDreyfus2	0.1122	0.0126	0.0884	0.2629	0.06
uGauss1	2.8942	8.3762	2.2345	9.7323	0.28
uGauss2	3.5399	12.5309	2.6525	13.2235	0.26
uGauss3	2.6559	7.0538	2.1042	7.8999	0.27
uNeuroOne	0.2830	0.0801	0.2312	0.5665	0.04

kable(apply(res[,,2,], c(3,1), min), caption = "BackpropBatch")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

BackpropBatch
	RMSE	MSE	MAE	WAE	time
mDette	1.3901	1.9323	1.1149	6.5646	5.86
mFriedman	0.0356	0.0013	0.0274	0.1387	5.97
mIshigami	2.6336	6.9360	2.2660	6.7254	7.58
mRef153	3.5771	12.7953	2.6208	13.7695	1.75
uDmod1	0.3276	0.1073	0.2773	0.7405	0.78
uDmod2	0.1722	0.0296	0.1346	0.4173	0.76
uDreyfus1	0.3256	0.1060	0.2575	0.6937	0.68
uDreyfus2	0.3019	0.0911	0.2342	0.8008	0.69
uGauss1	14.0515	197.4456	11.3434	38.7526	2.99
uGauss2	12.6808	160.8028	10.3220	26.8884	2.80
uGauss3	5.9202	35.0490	4.6467	16.3500	2.78
uNeuroOne	0.2885	0.0832	0.2358	0.6160	0.65

kable(apply(res[,,3,], c(3,1), min), caption = "BackpropChunk")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

BackpropChunk
	RMSE	MSE	MAE	WAE	time
mDette	0.6634	0.4401	0.5049	2.6444	0.59
mFriedman	0.0395	0.0016	0.0308	0.1718	0.62
mIshigami	1.8762	3.5200	1.3996	7.6244	0.72
mRef153	3.3911	11.4996	2.3498	12.7548	0.17
uDmod1	0.1306	0.0171	0.0888	0.4049	0.06
uDmod2	0.0799	0.0064	0.0631	0.1663	0.07
uDreyfus1	0.0717	0.0051	0.0590	0.1764	0.06
uDreyfus2	0.1175	0.0138	0.0930	0.2637	0.06
uGauss1	2.8405	8.0685	2.1772	8.4196	0.29
uGauss2	3.2770	10.7389	2.5878	11.2537	0.28
uGauss3	2.7491	7.5574	2.1950	7.3436	0.28
uNeuroOne	0.2869	0.0823	0.2385	0.5749	0.06

kable(apply(res[,,4,], c(3,1), min), caption = "BackpropMomentum")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

BackpropMomentum
	RMSE	MSE	MAE	WAE	time
mDette	0.6550	0.4291	0.4905	2.2000	0.57
mFriedman	0.0476	0.0023	0.0375	0.1649	0.59
mIshigami	2.6391	6.9649	2.0594	9.5926	0.70
mRef153	3.4035	11.5835	2.3942	11.7853	0.17
uDmod1	0.1223	0.0149	0.0962	0.3378	0.07
uDmod2	0.0681	0.0046	0.0570	0.1612	0.07
uDreyfus1	0.0801	0.0064	0.0613	0.2021	0.06
uDreyfus2	0.1177	0.0138	0.0931	0.2634	0.06
uGauss1	2.7202	7.3997	2.1139	8.0885	0.29
uGauss2	3.2581	10.6155	2.5245	10.3705	0.28
uGauss3	2.8889	8.3458	2.2397	7.7426	0.28
uNeuroOne	0.2840	0.0807	0.2325	0.5243	0.06

kable(apply(res[,,5,], c(3,1), min), caption = "BackpropWeightDecay")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

BackpropWeightDecay
	RMSE	MSE	MAE	WAE	time
mDette	0.6556	0.4298	0.5026	2.3936	0.58
mFriedman	0.0423	0.0018	0.0329	0.1724	0.61
mIshigami	2.6245	6.8878	2.0449	8.3924	0.75
mRef153	3.3903	11.4940	2.3312	13.5369	0.17
uDmod1	0.1398	0.0196	0.1026	0.4284	0.07
uDmod2	0.0840	0.0071	0.0661	0.2009	0.08
uDreyfus1	0.0793	0.0063	0.0604	0.1641	0.06
uDreyfus2	0.1174	0.0138	0.0928	0.2658	0.06
uGauss1	2.8753	8.2674	2.3058	8.5280	0.29
uGauss2	3.8043	14.4727	3.1540	10.1419	0.28
uGauss3	2.8868	8.3334	2.2915	8.1451	0.28
uNeuroOne	0.2856	0.0816	0.2353	0.5634	0.06

kable(apply(res[,,6,], c(3,1), min), caption = "Quickprop")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

Quickprop
	RMSE	MSE	MAE	WAE	time
mDette	7.1061	50.4966	5.7241	28.2385	6.46
mFriedman	0.1616	0.0261	0.1255	0.5029	6.50
mIshigami	3.4786	12.1008	2.8157	13.1137	8.72
mRef153	8.2933	68.7780	6.3301	23.9933	1.80
uDmod1	0.5187	0.2690	0.4371	1.0104	0.77
uDmod2	0.3034	0.0920	0.2507	0.7222	0.76
uDreyfus1	0.1885	0.0355	0.1319	0.5113	0.70
uDreyfus2	0.2410	0.0581	0.1938	0.5141	0.69
uGauss1	27.1904	739.3189	21.5229	66.1948	3.23
uGauss2	24.6889	609.5424	19.6622	51.2912	3.03
uGauss3	28.0694	787.8906	22.3771	52.4043	2.98
uNeuroOne	0.5304	0.2813	0.4235	1.2829	0.64

kable(apply(res[,,7,], c(3,1), min), caption = "SCG")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

SCG
	RMSE	MSE	MAE	WAE	time
mDette	0.4080	0.1664	0.3044	1.7469	0.88
mFriedman	0.0197	0.0004	0.0154	0.0616	0.80
mIshigami	0.6909	0.4774	0.5132	2.7162	1.18
mRef153	3.2673	10.6755	2.2672	13.6828	0.24
uDmod1	0.0547	0.0030	0.0436	0.1388	0.11
uDmod2	0.0523	0.0027	0.0437	0.1159	0.11
uDreyfus1	0.1002	0.0100	0.0815	0.2259	0.08
uDreyfus2	0.1041	0.0108	0.0826	0.2673	0.09
uGauss1	2.8599	8.1791	2.3042	8.8957	0.47
uGauss2	5.0971	25.9805	3.6184	14.6431	0.42
uGauss3	3.6281	13.1631	2.7798	10.6704	0.36
uNeuroOne	0.2866	0.0821	0.2342	0.6125	0.06

kable(apply(res[,,8,], c(3,1), min), caption = "Std_Backpropagation")%>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"))

Std_Backpropagation
	RMSE	MSE	MAE	WAE	time
mDette	0.4917	0.2418	0.3797	1.9109	0.56
mFriedman	0.0402	0.0016	0.0322	0.1328	0.59
mIshigami	0.9781	0.9566	0.7572	3.5234	0.69
mRef153	3.2695	10.6895	2.3606	13.0491	0.17
uDmod1	0.1106	0.0122	0.0871	0.3298	0.07
uDmod2	0.0757	0.0057	0.0626	0.1553	0.06
uDreyfus1	0.1077	0.0116	0.0871	0.2244	0.06
uDreyfus2	0.1289	0.0166	0.0998	0.2835	0.06
uGauss1	2.8593	8.1756	2.2813	8.6746	0.28
uGauss2	3.4239	11.7233	2.6003	13.0320	0.28
uGauss3	2.9484	8.6932	2.3186	7.7125	0.28
uNeuroOne	0.2834	0.0803	0.2296	0.5755	0.06

NNbenchmark Generic Function with RSNNS

All members of GSoC 2019 & 2020

2020/08/07

1 Setup

1.1 Packages and options

1.2 Datasets to test

1.3 RSNNS trainPredict arguments - inputs x, y

2 Launch package’s trainPredict

3 Results