R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> # 1. vdp.mixt: moodien loytyminen eri dimensiolla, naytemaarilla ja komponenteilla
> #   -> ainakin nopea check
> 
> #######################################################################
> 
> # Generate random data from five Gaussians. 
> # Detect modes with vdp-gm. 
> # Plot data points and detected clusters with variance ellipses
> 
> #######################################################################
> 
> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> #source("~/Rpackages/netresponse/netresponse/R/detect.responses.R")
> #source("~/Rpackages/netresponse/netresponse/R/internals.R")
> #source("~/Rpackages/netresponse/netresponse/R/vdp.mixt.R")
> #dyn.load("/home/tuli/Rpackages/netresponse/netresponse/src/netresponse.so")
> 
> #########  Generate DATA #############################################
> 
> # Generate Nc components from normal-inverseGamma prior
> 
> set.seed(12346)
> 
> dd <- 3   # Dimensionality of data
> Nc <- 5   # Number of components
> Ns <- 200 # Number of data points
> sd0 <- 3  # component spread
> rgam.shape = 2 # parameters for Gamma distribution 
> rgam.scale = 2 # parameters for Gamma distribution to define precisions
> 
> 
> # Generate means and variances (covariance diagonals) for the components 
> component.means <- matrix(rnorm(Nc*dd, mean = 0, sd = sd0), nrow = Nc, ncol = dd)
> component.vars <- matrix(1/rgamma(Nc*dd, shape = rgam.shape, scale = rgam.scale), 
+ 	                 nrow = Nc, ncol = dd)
> component.sds <- sqrt(component.vars)
> 
> 
> # Size for each component -> sample randomly for each data point from uniform distr.
> # i.e. cluster assignments
> sample2comp <- sample.int(Nc, Ns, replace = TRUE)
> 
> D <- array(NA, dim = c(Ns, dd))
> for (i in 1:Ns)  {
+     # component identity of this sample
+     ci <- sample2comp[[i]]
+     cm <- component.means[ci,]
+     csd <- component.sds[ci,]
+     D[i,] <- rnorm(dd, mean = cm, sd = csd)
+ }
> 
> 
> ######################################################################
> 
> # Fit mixture model
> out <- mixture.model(D, mixture.method = "bic")
> 
> # FIXME rowmeans(qofz) is constant but not 1
> #qofz <- P.r.s(t(D), list(mu = out$mu, sd = out$sd, w = out$w), log = FALSE)
> 
> ############################################################
> 
> # Compare input data and results
> 
> ord.out <- order(out$mu[,1])
> ord.in <- order(component.means[,1])
> 
> means.out <- out$mu[ord.out,]
> means.in <- component.means[ord.in,]
> 
> # Cluster stds and variances
> sds.out <- out$sd[ord.out,]
> sds.in  <- sqrt(component.vars[ord.in,])
> 
> # -----------------------------------------------------------
> 
> vars.out <- sds.out^2
> vars.in <- sds.in^2
> 
> # Check correspondence between input and output
> if (length(means.in) == length(means.out)) {
+    cm <- cor(as.vector(means.in), as.vector(means.out))
+    csd <- cor(as.vector(sds.in), as.vector(sds.out))
+ }
> 
> # Plot results (assuming 2D)
> 
> ran <- range(c(as.vector(means.in - 2*vars.in), 
+                as.vector(means.in + 2*vars.in), 
+ 	       as.vector(means.out + 2*vars.out), 
+ 	       as.vector(means.out - 2*vars.out)))
> 
> plot(D, pch = 20, main = paste("Cor.means:", round(cm,3), "/ Cor.sds:", round(csd,3)), xlim = ran, ylim = ran) 
> for (ci in 1:nrow(means.out))  { add.ellipse(centroid = means.out[ci,], covmat = diag(vars.out[ci,]), col = "red") }
> for (ci in 1:nrow(means.in))  { add.ellipse(centroid = means.in[ci,], covmat = diag(vars.in[ci,]), col = "blue") }
> 
> ######################################################
> 
> #for (ci in 1:nrow(means.out))  {
> #    points(means.out[ci,1], means.out[ci,2], col = "red", pch = 19)
> #    el <- ellipse(matrix(c(vars.out[ci,1],0,0,vars.out[ci,2]),2), centre = means.out[ci,])
> #    lines(el, col = "red") 						  
> #}
> 
> #for (ci in 1:nrow(means.in))  {
> #    points(means.in[ci,1], means.in[ci,2], col = "blue", pch = 19)
> #    el <- ellipse(matrix(c(vars.in[ci,1],0,0,vars.in[ci,2]),2), centre = means.in[ci,])
> #    lines(el, col = "blue") 						  
> #}
> 
> 
> 
> 
> 
> 
> proc.time()
   user  system elapsed 
   4.45    0.35    4.79 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> # 1. vdp.mixt: moodien loytyminen eri dimensiolla, naytemaarilla ja komponenteilla
> #   -> ainakin nopea check
> 
> #######################################################################
> 
> # Generate random data from five Gaussians. 
> # Detect modes with vdp-gm. 
> # Plot data points and detected clusters with variance ellipses
> 
> #######################################################################
> 
> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> #source("~/Rpackages/netresponse/netresponse/R/detect.responses.R")
> #source("~/Rpackages/netresponse/netresponse/R/internals.R")
> #source("~/Rpackages/netresponse/netresponse/R/vdp.mixt.R")
> #dyn.load("/home/tuli/Rpackages/netresponse/netresponse/src/netresponse.so")
> 
> #########  Generate DATA #############################################
> 
> # Generate Nc components from normal-inverseGamma prior
> 
> set.seed(12346)
> 
> dd <- 3   # Dimensionality of data
> Nc <- 5   # Number of components
> Ns <- 200 # Number of data points
> sd0 <- 3  # component spread
> rgam.shape = 2 # parameters for Gamma distribution 
> rgam.scale = 2 # parameters for Gamma distribution to define precisions
> 
> 
> # Generate means and variances (covariance diagonals) for the components 
> component.means <- matrix(rnorm(Nc*dd, mean = 0, sd = sd0), nrow = Nc, ncol = dd)
> component.vars <- matrix(1/rgamma(Nc*dd, shape = rgam.shape, scale = rgam.scale), 
+ 	                 nrow = Nc, ncol = dd)
> component.sds <- sqrt(component.vars)
> 
> 
> # Size for each component -> sample randomly for each data point from uniform distr.
> # i.e. cluster assignments
> sample2comp <- sample.int(Nc, Ns, replace = TRUE)
> 
> D <- array(NA, dim = c(Ns, dd))
> for (i in 1:Ns)  {
+     # component identity of this sample
+     ci <- sample2comp[[i]]
+     cm <- component.means[ci,]
+     csd <- component.sds[ci,]
+     D[i,] <- rnorm(dd, mean = cm, sd = csd)
+ }
> 
> 
> ######################################################################
> 
> # Fit mixture model
> out <- mixture.model(D, mixture.method = "bic")
> 
> # FIXME rowmeans(qofz) is constant but not 1
> #qofz <- P.r.s(t(D), list(mu = out$mu, sd = out$sd, w = out$w), log = FALSE)
> 
> ############################################################
> 
> # Compare input data and results
> 
> ord.out <- order(out$mu[,1])
> ord.in <- order(component.means[,1])
> 
> means.out <- out$mu[ord.out,]
> means.in <- component.means[ord.in,]
> 
> # Cluster stds and variances
> sds.out <- out$sd[ord.out,]
> sds.in  <- sqrt(component.vars[ord.in,])
> 
> # -----------------------------------------------------------
> 
> vars.out <- sds.out^2
> vars.in <- sds.in^2
> 
> # Check correspondence between input and output
> if (length(means.in) == length(means.out)) {
+    cm <- cor(as.vector(means.in), as.vector(means.out))
+    csd <- cor(as.vector(sds.in), as.vector(sds.out))
+ }
> 
> # Plot results (assuming 2D)
> 
> ran <- range(c(as.vector(means.in - 2*vars.in), 
+                as.vector(means.in + 2*vars.in), 
+ 	       as.vector(means.out + 2*vars.out), 
+ 	       as.vector(means.out - 2*vars.out)))
> 
> plot(D, pch = 20, main = paste("Cor.means:", round(cm,3), "/ Cor.sds:", round(csd,3)), xlim = ran, ylim = ran) 
> for (ci in 1:nrow(means.out))  { add.ellipse(centroid = means.out[ci,], covmat = diag(vars.out[ci,]), col = "red") }
> for (ci in 1:nrow(means.in))  { add.ellipse(centroid = means.in[ci,], covmat = diag(vars.in[ci,]), col = "blue") }
> 
> ######################################################
> 
> #for (ci in 1:nrow(means.out))  {
> #    points(means.out[ci,1], means.out[ci,2], col = "red", pch = 19)
> #    el <- ellipse(matrix(c(vars.out[ci,1],0,0,vars.out[ci,2]),2), centre = means.out[ci,])
> #    lines(el, col = "red") 						  
> #}
> 
> #for (ci in 1:nrow(means.in))  {
> #    points(means.in[ci,1], means.in[ci,2], col = "blue", pch = 19)
> #    el <- ellipse(matrix(c(vars.in[ci,1],0,0,vars.in[ci,2]),2), centre = means.in[ci,])
> #    lines(el, col = "blue") 						  
> #}
> 
> 
> 
> 
> 
> 
> proc.time()
   user  system elapsed 
   3.59    0.29    3.87 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> # Validate mixture models
> 
> # Generate random data from five Gaussians. 
> # Detect modes 
> # Plot data points and detected clusters 
> 
> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> 
> #fs <- list.files("~/Rpackages/netresponse/netresponse/R/", full.names = TRUE); for (f in fs) {source(f)}; dyn.load("/home/tuli/Rpackages/netresponse/netresponse/src/netresponse.so")
> 
> #########  Generate DATA #######################
> 
> res <- generate.toydata()
> D <- res$data
> component.means <- res$means
> component.sds   <- res$sds
> sample2comp     <- res$sample2comp
> 
> ######################################################################
> 
> par(mfrow = c(2,1))
> 
> for (mm in c("vdp", "bic")) {
+ 
+   # Fit nonparametric Gaussian mixture model
+   #source("~/Rpackages/netresponse/netresponse/R/vdp.mixt.R")
+   out <- mixture.model(D, mixture.method = mm, max.responses = 10, pca.basis = FALSE)
+ 
+   ############################################################
+ 
+   # Compare input data and results
+ 
+   ord.out <- order(out$mu[,1])
+   ord.in <- order(component.means[,1])
+ 
+   means.out <- out$mu[ord.out,]
+   means.in <- component.means[ord.in,]
+ 
+   # Cluster stds and variances
+   sds.out <- out$sd[ord.out,]
+   vars.out <- sds.out^2
+ 
+   sds.in  <- component.sds[ord.in,]
+   vars.in <- sds.in^2
+ 
+   # Check correspondence between input and output
+   if (length(means.in) == length(means.out)) {
+     cm <- cor(as.vector(means.in), as.vector(means.out))
+     csd <- cor(as.vector(sds.in), as.vector(sds.out))
+   }
+ 
+   # Plot results (assuming 2D)
+   ran <- range(c(as.vector(means.in - 2*vars.in), 
+                as.vector(means.in + 2*vars.in), 
+ 	       as.vector(means.out + 2*vars.out), 
+ 	       as.vector(means.out - 2*vars.out)))
+ 
+   real.modes <- sample2comp
+   obs.modes <- apply(out$qofz, 1, which.max)
+ 
+   # plot(D, pch = 20, main = paste(mm, "/ cor.means:", round(cm,6), "/ Cor.sds:", round(csd,6)), xlim = ran, ylim = ran)
+   
+   # plot(D, pch = real.modes, col = obs.modes, main = paste(mm, "/ cor.means:", round(cm,6), "/ Cor.sds:", round(csd,6)), xlim = ran, ylim = ran)
+   
+   # for (ci in 1:nrow(means.out))  { add.ellipse(centroid = means.out[ci,], covmat = diag(vars.out[ci,]), col = "red") }
+   # for (ci in 1:nrow(means.in))  { add.ellipse(centroid = means.in[ci,], covmat = diag(vars.in[ci,]), col = "blue") }
+ 
+ }
> 
> 
> proc.time()
   user  system elapsed 
   4.50    0.50    4.98 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> # Validate mixture models
> 
> # Generate random data from five Gaussians. 
> # Detect modes 
> # Plot data points and detected clusters 
> 
> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> 
> #fs <- list.files("~/Rpackages/netresponse/netresponse/R/", full.names = TRUE); for (f in fs) {source(f)}; dyn.load("/home/tuli/Rpackages/netresponse/netresponse/src/netresponse.so")
> 
> #########  Generate DATA #######################
> 
> res <- generate.toydata()
> D <- res$data
> component.means <- res$means
> component.sds   <- res$sds
> sample2comp     <- res$sample2comp
> 
> ######################################################################
> 
> par(mfrow = c(2,1))
> 
> for (mm in c("vdp", "bic")) {
+ 
+   # Fit nonparametric Gaussian mixture model
+   #source("~/Rpackages/netresponse/netresponse/R/vdp.mixt.R")
+   out <- mixture.model(D, mixture.method = mm, max.responses = 10, pca.basis = FALSE)
+ 
+   ############################################################
+ 
+   # Compare input data and results
+ 
+   ord.out <- order(out$mu[,1])
+   ord.in <- order(component.means[,1])
+ 
+   means.out <- out$mu[ord.out,]
+   means.in <- component.means[ord.in,]
+ 
+   # Cluster stds and variances
+   sds.out <- out$sd[ord.out,]
+   vars.out <- sds.out^2
+ 
+   sds.in  <- component.sds[ord.in,]
+   vars.in <- sds.in^2
+ 
+   # Check correspondence between input and output
+   if (length(means.in) == length(means.out)) {
+     cm <- cor(as.vector(means.in), as.vector(means.out))
+     csd <- cor(as.vector(sds.in), as.vector(sds.out))
+   }
+ 
+   # Plot results (assuming 2D)
+   ran <- range(c(as.vector(means.in - 2*vars.in), 
+                as.vector(means.in + 2*vars.in), 
+ 	       as.vector(means.out + 2*vars.out), 
+ 	       as.vector(means.out - 2*vars.out)))
+ 
+   real.modes <- sample2comp
+   obs.modes <- apply(out$qofz, 1, which.max)
+ 
+   # plot(D, pch = 20, main = paste(mm, "/ cor.means:", round(cm,6), "/ Cor.sds:", round(csd,6)), xlim = ran, ylim = ran)
+   
+   # plot(D, pch = real.modes, col = obs.modes, main = paste(mm, "/ cor.means:", round(cm,6), "/ Cor.sds:", round(csd,6)), xlim = ran, ylim = ran)
+   
+   # for (ci in 1:nrow(means.out))  { add.ellipse(centroid = means.out[ci,], covmat = diag(vars.out[ci,]), col = "red") }
+   # for (ci in 1:nrow(means.in))  { add.ellipse(centroid = means.in[ci,], covmat = diag(vars.in[ci,]), col = "blue") }
+ 
+ }
> 
> 
> proc.time()
   user  system elapsed 
   3.35    0.21    3.56 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> 
> # Three MODES
> 
> # set.seed(34884)
> set.seed(3488400)
> 
> Ns <- 200
> Nd <- 2
> 
> D3 <- rbind(matrix(rnorm(Ns*Nd, mean = 0), ncol = Nd), 
+       	    matrix(rnorm(Ns*Nd, mean = 3), ncol = Nd),
+       	    cbind(rnorm(Ns, mean = -3), rnorm(Ns, mean = 3))
+ 	    )
> 
> #X11()
> par(mfrow = c(2,2))
> for (mm in c("vdp", "bic")) {
+   for (pp in c(FALSE, TRUE)) {
+ 
+     # Fit nonparametric Gaussian mixture model
+     out <- mixture.model(D3, mixture.method = mm, pca.basis = pp)
+     plot(D3, col = apply(out$qofz, 1, which.max), main = paste(mm, "/ pca:",  pp)) 
+ 
+   }
+ }
> 
> # VDP is less sensitive than BIC in detecting Gaussian modes (more
> # separation between the clusters needed)
> 
> # pca.basis option is less important for sensitive detection but
> # it will help to avoid overfitting to unimodal features that
> # are not parallel to the axes (unimodal distribution often becomes
> # splitted in two or more clusters in these cases)
> 
> 
> proc.time()
   user  system elapsed 
   7.20    0.46    7.65 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> 
> # Three MODES
> 
> # set.seed(34884)
> set.seed(3488400)
> 
> Ns <- 200
> Nd <- 2
> 
> D3 <- rbind(matrix(rnorm(Ns*Nd, mean = 0), ncol = Nd), 
+       	    matrix(rnorm(Ns*Nd, mean = 3), ncol = Nd),
+       	    cbind(rnorm(Ns, mean = -3), rnorm(Ns, mean = 3))
+ 	    )
> 
> #X11()
> par(mfrow = c(2,2))
> for (mm in c("vdp", "bic")) {
+   for (pp in c(FALSE, TRUE)) {
+ 
+     # Fit nonparametric Gaussian mixture model
+     out <- mixture.model(D3, mixture.method = mm, pca.basis = pp)
+     plot(D3, col = apply(out$qofz, 1, which.max), main = paste(mm, "/ pca:",  pp)) 
+ 
+   }
+ }
> 
> # VDP is less sensitive than BIC in detecting Gaussian modes (more
> # separation between the clusters needed)
> 
> # pca.basis option is less important for sensitive detection but
> # it will help to avoid overfitting to unimodal features that
> # are not parallel to the axes (unimodal distribution often becomes
> # splitted in two or more clusters in these cases)
> 
> 
> proc.time()
   user  system elapsed 
   5.43    0.25    5.65 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> skip <- FALSE
> 
> if (!skip) {
+ 
+ library(netresponse)
+ 
+ # SINGLE MODE
+ 
+ # Produce test data that has full covariance
+ # It is expected that
+ # pca.basis = FALSE splits Gaussian with full covariance into two modes
+ # pca.basis = TRUE should detect just a single mode
+ 
+ Ns <- 200
+ Nd <- 2
+ k <- 1.5
+ 
+ D2 <- matrix(rnorm(Ns*Nd), ncol = Nd) %*% rbind(c(1,k), c(k,1))
+ 
+ par(mfrow = c(2,2))
+ for (mm in c("vdp", "bic")) {
+   for (pp in c(FALSE, TRUE)) {
+ 
+     # Fit nonparametric Gaussian mixture model
+     out <- mixture.model(D2, mixture.method = mm, pca.basis = pp)
+     plot(D2, col = apply(out$qofz, 1, which.max), main = paste("mm:" , mm, "/ pp:",  pp)) 
+ 
+   }
+ }
+ 
+ }
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> 
> proc.time()
   user  system elapsed 
   3.81    0.35    4.15 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> skip <- FALSE
> 
> if (!skip) {
+ 
+ library(netresponse)
+ 
+ # SINGLE MODE
+ 
+ # Produce test data that has full covariance
+ # It is expected that
+ # pca.basis = FALSE splits Gaussian with full covariance into two modes
+ # pca.basis = TRUE should detect just a single mode
+ 
+ Ns <- 200
+ Nd <- 2
+ k <- 1.5
+ 
+ D2 <- matrix(rnorm(Ns*Nd), ncol = Nd) %*% rbind(c(1,k), c(k,1))
+ 
+ par(mfrow = c(2,2))
+ for (mm in c("vdp", "bic")) {
+   for (pp in c(FALSE, TRUE)) {
+ 
+     # Fit nonparametric Gaussian mixture model
+     out <- mixture.model(D2, mixture.method = mm, pca.basis = pp)
+     plot(D2, col = apply(out$qofz, 1, which.max), main = paste("mm:" , mm, "/ pp:",  pp)) 
+ 
+   }
+ }
+ 
+ }
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> 
> proc.time()
   user  system elapsed 
   4.92    0.32    5.23 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> # Play with different options and check their effect on  running times for bic and vdp 
> 
> skip <- TRUE
> 
> if (!skip) {
+ 
+   Ns <- 100
+   Nd <- 2
+ 
+   set.seed(3488400)
+ 
+   D <- cbind(
+ 
+      	rbind(matrix(rnorm(Ns*Nd, mean = 0), ncol = Nd), 
+        	      matrix(rnorm(Ns*Nd, mean = 2), ncol = Nd),
+       	      cbind(rnorm(Ns, mean = -1), rnorm(Ns, mean = 3))
+  	    ), 
+ 
+      	rbind(matrix(rnorm(Ns*Nd, mean = 0), ncol = Nd), 
+        	      matrix(rnorm(Ns*Nd, mean = 2), ncol = Nd),
+       	      cbind(rnorm(Ns, mean = -1), rnorm(Ns, mean = 3))
+  	    )
+ 	    )
+ 
+   rownames(D) <- paste("R", 1:nrow(D), sep = "-")
+   colnames(D) <- paste("C", 1:ncol(D), sep = "-")
+ 
+   ts <- c()
+   for (mm in c("bic", "vdp")) {
+ 
+ 
+     # NOTE: no PCA basis needed with mixture.method = "bic"
+     tt <- system.time(detect.responses(D, verbose = TRUE, max.responses = 5, 
+ 	   		       mixture.method = mm, information.criterion = "BIC", 
+ 			       merging.threshold = 0, bic.threshold = 0, pca.basis = TRUE))
+ 
+     print(paste(mm, ":", round(tt[["elapsed"]], 3)))
+     ts[[mm]] <- tt[["elapsed"]]
+   }
+ 
+    print(paste(names(ts)[[1]], "/", names(ts)[[2]], ": ", round(ts[[1]]/ts[[2]], 3)))
+ 
+ }
> 
> # -> VDP is much faster when sample sizes increase 
> # 1000 samples -> 25-fold speedup with VDP
> 
> 
> 
> proc.time()
   user  system elapsed 
   0.23    0.04    0.26 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> # Play with different options and check their effect on  running times for bic and vdp 
> 
> skip <- TRUE
> 
> if (!skip) {
+ 
+   Ns <- 100
+   Nd <- 2
+ 
+   set.seed(3488400)
+ 
+   D <- cbind(
+ 
+      	rbind(matrix(rnorm(Ns*Nd, mean = 0), ncol = Nd), 
+        	      matrix(rnorm(Ns*Nd, mean = 2), ncol = Nd),
+       	      cbind(rnorm(Ns, mean = -1), rnorm(Ns, mean = 3))
+  	    ), 
+ 
+      	rbind(matrix(rnorm(Ns*Nd, mean = 0), ncol = Nd), 
+        	      matrix(rnorm(Ns*Nd, mean = 2), ncol = Nd),
+       	      cbind(rnorm(Ns, mean = -1), rnorm(Ns, mean = 3))
+  	    )
+ 	    )
+ 
+   rownames(D) <- paste("R", 1:nrow(D), sep = "-")
+   colnames(D) <- paste("C", 1:ncol(D), sep = "-")
+ 
+   ts <- c()
+   for (mm in c("bic", "vdp")) {
+ 
+ 
+     # NOTE: no PCA basis needed with mixture.method = "bic"
+     tt <- system.time(detect.responses(D, verbose = TRUE, max.responses = 5, 
+ 	   		       mixture.method = mm, information.criterion = "BIC", 
+ 			       merging.threshold = 0, bic.threshold = 0, pca.basis = TRUE))
+ 
+     print(paste(mm, ":", round(tt[["elapsed"]], 3)))
+     ts[[mm]] <- tt[["elapsed"]]
+   }
+ 
+    print(paste(names(ts)[[1]], "/", names(ts)[[2]], ": ", round(ts[[1]]/ts[[2]], 3)))
+ 
+ }
> 
> # -> VDP is much faster when sample sizes increase 
> # 1000 samples -> 25-fold speedup with VDP
> 
> 
> 
> proc.time()
   user  system elapsed 
   0.34    0.06    0.39 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> # Generate Nc components from normal-inverseGamma prior
> 
> set.seed(12346)
> 
> Ns <- 300
> Nd <- 2
> 
> # Isotropic cloud
> D1 <- matrix(rnorm(Ns*Nd), ncol = Nd) 
> 
> # Single diagonal mode
> D2 <- matrix(rnorm(Ns*Nd), ncol = Nd) %*% rbind(c(1,2), c(2,1)) 
> 
> # Two isotropic modes
> D3 <- rbind(matrix(rnorm(Ns/2*Nd), ncol = Nd), matrix(rnorm(Ns/2*Nd, mean = 3), ncol = Nd))
> D <- cbind(D1, D2, D3)
> 
> colnames(D) <- paste("Feature-",  1:ncol(D), sep = "")
> rownames(D) <- paste("Sample-", 1:nrow(D), sep = "")
> 
> 
> proc.time()
   user  system elapsed 
   0.21    0.01    0.21 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> # Generate Nc components from normal-inverseGamma prior
> 
> set.seed(12346)
> 
> Ns <- 300
> Nd <- 2
> 
> # Isotropic cloud
> D1 <- matrix(rnorm(Ns*Nd), ncol = Nd) 
> 
> # Single diagonal mode
> D2 <- matrix(rnorm(Ns*Nd), ncol = Nd) %*% rbind(c(1,2), c(2,1)) 
> 
> # Two isotropic modes
> D3 <- rbind(matrix(rnorm(Ns/2*Nd), ncol = Nd), matrix(rnorm(Ns/2*Nd, mean = 3), ncol = Nd))
> D <- cbind(D1, D2, D3)
> 
> colnames(D) <- paste("Feature-",  1:ncol(D), sep = "")
> rownames(D) <- paste("Sample-", 1:nrow(D), sep = "")
> 
> 
> proc.time()
   user  system elapsed 
   0.21    0.04    0.26 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> skip <- TRUE
> 
> if (!skip) {
+ 
+ # 2. netresponse test
+ # test later with varying parameters
+ 
+ # Load the package
+ library(netresponse)
+ #load("../data/toydata.rda")
+ fs <- list.files("../R/", full.names = TRUE); for (f in fs) {source(f)};
+ 
+ data(toydata)
+ 
+ D <- toydata$emat
+ netw <- toydata$netw
+ 
+ # The toy data is random data with 10 features (genes). 
+ # The features 
+ rf <- c(4, 5, 6)
+ #form a subnetwork with coherent responses
+ # with means 
+ r1 <- c(0, 3, 0)
+ r2 <- c(-5, 0, 2)
+ r3 <- c(5, -3, -3)
+ mu.real <- rbind(r1, r2, r3)
+ # real weights
+ w.real <- c(70, 70, 60)/200
+ # and unit variances
+ rv <- 1
+ 
+ # Fit the model
+ #res <- detect.responses(D, netw, verbose = TRUE, mc.cores = 2)
+ #res <- detect.responses(D, netw, verbose = TRUE, max.responses = 4)
+ 
+ res <- detect.responses(D, netw, verbose = TRUE, max.responses = 3, mixture.method = "bic", information.criterion = "BIC", merging.threshold = 1, bic.threshold = 10, pca.basis = FALSE)
+ 
+ print("OK")
+ 
+ # Subnets (each is a list of nodes)
+ subnets <- get.subnets(res)
+ 
+ # the correct subnet is retrieved in subnet number 2:
+ #> subnet[[2]]
+ #[1] "feat4" "feat5" "feat6"
+ 
+ # how about responses
+ # Retrieve model for the subnetwork with lowest cost function value
+ # means, standard devations and weights for the components
+ if (!is.null(subnets)) {
+ m <- get.model.parameters(res, subnet.id = "Subnet-2")
+ 
+ # order retrieved and real response means by the first feature 
+ # (to ensure responses are listed in the same order)
+ # and compare deviation from correct solution
+ ord.obs <- order(m$mu[,1])
+ ord.real <- order(mu.real[,1])
+ 
+ print(paste("Correlation between real and observed responses:", cor(as.vector(m$mu[ord.obs,]), as.vector(mu.real[ord.real,]))))
+ 
+ # all real variances are 1, compare to observed ones
+ print(paste("Maximum deviation from real variances: ", max(abs(rv - range(m$sd))/rv)))
+ 
+ # weights deviate somewhat, this is likely due to relatively small sample size
+ #print("Maximum deviation from real weights: ")
+ #print( (w.real[ord.real] - m$w[ord.obs])/w.real[ord.real])
+ 
+ print("estimated and real mean matrices")
+ print(m$mu[ord.obs,])
+ print(mu.real[ord.real,])
+ 
+ }
+ 
+ }
> 
> proc.time()
   user  system elapsed 
   0.17    0.04    0.20 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> skip <- TRUE
> 
> if (!skip) {
+ 
+ # 2. netresponse test
+ # test later with varying parameters
+ 
+ # Load the package
+ library(netresponse)
+ #load("../data/toydata.rda")
+ fs <- list.files("../R/", full.names = TRUE); for (f in fs) {source(f)};
+ 
+ data(toydata)
+ 
+ D <- toydata$emat
+ netw <- toydata$netw
+ 
+ # The toy data is random data with 10 features (genes). 
+ # The features 
+ rf <- c(4, 5, 6)
+ #form a subnetwork with coherent responses
+ # with means 
+ r1 <- c(0, 3, 0)
+ r2 <- c(-5, 0, 2)
+ r3 <- c(5, -3, -3)
+ mu.real <- rbind(r1, r2, r3)
+ # real weights
+ w.real <- c(70, 70, 60)/200
+ # and unit variances
+ rv <- 1
+ 
+ # Fit the model
+ #res <- detect.responses(D, netw, verbose = TRUE, mc.cores = 2)
+ #res <- detect.responses(D, netw, verbose = TRUE, max.responses = 4)
+ 
+ res <- detect.responses(D, netw, verbose = TRUE, max.responses = 3, mixture.method = "bic", information.criterion = "BIC", merging.threshold = 1, bic.threshold = 10, pca.basis = FALSE)
+ 
+ print("OK")
+ 
+ # Subnets (each is a list of nodes)
+ subnets <- get.subnets(res)
+ 
+ # the correct subnet is retrieved in subnet number 2:
+ #> subnet[[2]]
+ #[1] "feat4" "feat5" "feat6"
+ 
+ # how about responses
+ # Retrieve model for the subnetwork with lowest cost function value
+ # means, standard devations and weights for the components
+ if (!is.null(subnets)) {
+ m <- get.model.parameters(res, subnet.id = "Subnet-2")
+ 
+ # order retrieved and real response means by the first feature 
+ # (to ensure responses are listed in the same order)
+ # and compare deviation from correct solution
+ ord.obs <- order(m$mu[,1])
+ ord.real <- order(mu.real[,1])
+ 
+ print(paste("Correlation between real and observed responses:", cor(as.vector(m$mu[ord.obs,]), as.vector(mu.real[ord.real,]))))
+ 
+ # all real variances are 1, compare to observed ones
+ print(paste("Maximum deviation from real variances: ", max(abs(rv - range(m$sd))/rv)))
+ 
+ # weights deviate somewhat, this is likely due to relatively small sample size
+ #print("Maximum deviation from real weights: ")
+ #print( (w.real[ord.real] - m$w[ord.obs])/w.real[ord.real])
+ 
+ print("estimated and real mean matrices")
+ print(m$mu[ord.obs,])
+ print(mu.real[ord.real,])
+ 
+ }
+ 
+ }
> 
> proc.time()
   user  system elapsed 
   0.26    0.04    0.29 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> skip <- FALSE
> 
> if (!skip) {
+ # Visualization
+ 
+ library(netresponse)
+ 
+ #fs <- list.files("~/Rpackages/netresponse/netresponse/R/", full.names = T); for (f in fs) {source(f)}
+ 
+ source("toydata2.R")
+ 
+ # --------------------------------------------------------------------
+ 
+ set.seed(4243)
+ mixture.method <- "bic"
+ 
+ # --------------------------------------------------------------------
+ 
+ res <- detect.responses(D, verbose = TRUE, max.responses = 10, 
+ 	   		       mixture.method = mixture.method, information.criterion = "BIC", 
+ 			       merging.threshold = 1, bic.threshold = 10, pca.basis = FALSE)
+ 
+ res.pca <- detect.responses(D, verbose = TRUE, max.responses = 10, mixture.method = mixture.method, information.criterion = "BIC", merging.threshold = 1, bic.threshold = 10, pca.basis = TRUE)
+ 
+ # --------------------------------------------------------------------
+ 
+ k <- 1
+ 
+ # Incorrect VDP: two modes detected
+ # Correct BIC: single mode detected
+ subnet.id <- names(get.subnets(res))[[k]]
+ 
+ # Correct: single mode detected (VDP & BIC)
+ subnet.id.pca <- names(get.subnets(res.pca))[[k]]
+ 
+ # --------------------------------------------------------------------------------------------------
+ 
+ vis1 <- plot_responses(res, subnet.id, plot_mode = "pca", main = paste("NoPCA; NoDM"))
+ vis2 <- plot_responses(res, subnet.id, plot_mode = "pca", datamatrix = D, main = "NoPCA, DM")
+ vis3 <- plot_responses(res.pca, subnet.id.pca, plot_mode = "pca", main = "PCA, NoDM")
+ vis4 <- plot_responses(res.pca, subnet.id.pca, plot_mode = "pca", datamatrix = D, main = "PCA, DM")
+ 
+ # With original data: VDP overlearns; BIC works; with full covariance data 
+ # With PCA basis: modes detected ok with both VDP and BIC.
+ 
+ # ------------------------------------------------------------------------
+ 
+ # TODO
+ # pca.plot(res, subnet.id)
+ # plot_subnet(res, subnet.id) 
+ }
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
No network provided in function call: assuming fully connected nodes.
convert the network into edge matrix
removing self-links
matching the features between network and datamatrix
Filter the network to only keep the edges with highest mutual information
1 / 5
2 / 5
3 / 5
4 / 5
5 / 5
Compute cost for each variable
Computing model for node 1 / 6
Computing model for node 2 / 6
Computing model for node 3 / 6
Computing model for node 4 / 6
Computing model for node 5 / 6
Computing model for node 6 / 6
independent models done
Computing delta values for edge  1 / 15 

Computing delta values for edge  2 / 15 

Computing delta values for edge  3 / 15 

Computing delta values for edge  4 / 15 

Computing delta values for edge  5 / 15 

Computing delta values for edge  6 / 15 

Computing delta values for edge  7 / 15 

Computing delta values for edge  8 / 15 

Computing delta values for edge  9 / 15 

Computing delta values for edge  10 / 15 

Computing delta values for edge  11 / 15 

Computing delta values for edge  12 / 15 

Computing delta values for edge  13 / 15 

Computing delta values for edge  14 / 15 

Computing delta values for edge  15 / 15 

Combining groups,  6  group(s) left...

Combining groups,  5  group(s) left...

Combining groups,  4  group(s) left...

Combining groups,  3  group(s) left...

No network provided in function call: assuming fully connected nodes.
convert the network into edge matrix
removing self-links
matching the features between network and datamatrix
Filter the network to only keep the edges with highest mutual information
1 / 5
2 / 5
3 / 5
4 / 5
5 / 5
Compute cost for each variable
Computing model for node 1 / 6
Computing model for node 2 / 6
Computing model for node 3 / 6
Computing model for node 4 / 6
Computing model for node 5 / 6
Computing model for node 6 / 6
independent models done
Computing delta values for edge  1 / 15 

Computing delta values for edge  2 / 15 

Computing delta values for edge  3 / 15 

Computing delta values for edge  4 / 15 

Computing delta values for edge  5 / 15 

Computing delta values for edge  6 / 15 

Computing delta values for edge  7 / 15 

Computing delta values for edge  8 / 15 

Computing delta values for edge  9 / 15 

Computing delta values for edge  10 / 15 

Computing delta values for edge  11 / 15 

Computing delta values for edge  12 / 15 

Computing delta values for edge  13 / 15 

Computing delta values for edge  14 / 15 

Computing delta values for edge  15 / 15 

Combining groups,  6  group(s) left...

Combining groups,  5  group(s) left...

Combining groups,  4  group(s) left...

Combining groups,  3  group(s) left...

> 
> proc.time()
   user  system elapsed 
  30.54    0.35   30.89 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> skip <- FALSE
> 
> if (!skip) {
+ # Visualization
+ 
+ library(netresponse)
+ 
+ #fs <- list.files("~/Rpackages/netresponse/netresponse/R/", full.names = T); for (f in fs) {source(f)}
+ 
+ source("toydata2.R")
+ 
+ # --------------------------------------------------------------------
+ 
+ set.seed(4243)
+ mixture.method <- "bic"
+ 
+ # --------------------------------------------------------------------
+ 
+ res <- detect.responses(D, verbose = TRUE, max.responses = 10, 
+ 	   		       mixture.method = mixture.method, information.criterion = "BIC", 
+ 			       merging.threshold = 1, bic.threshold = 10, pca.basis = FALSE)
+ 
+ res.pca <- detect.responses(D, verbose = TRUE, max.responses = 10, mixture.method = mixture.method, information.criterion = "BIC", merging.threshold = 1, bic.threshold = 10, pca.basis = TRUE)
+ 
+ # --------------------------------------------------------------------
+ 
+ k <- 1
+ 
+ # Incorrect VDP: two modes detected
+ # Correct BIC: single mode detected
+ subnet.id <- names(get.subnets(res))[[k]]
+ 
+ # Correct: single mode detected (VDP & BIC)
+ subnet.id.pca <- names(get.subnets(res.pca))[[k]]
+ 
+ # --------------------------------------------------------------------------------------------------
+ 
+ vis1 <- plot_responses(res, subnet.id, plot_mode = "pca", main = paste("NoPCA; NoDM"))
+ vis2 <- plot_responses(res, subnet.id, plot_mode = "pca", datamatrix = D, main = "NoPCA, DM")
+ vis3 <- plot_responses(res.pca, subnet.id.pca, plot_mode = "pca", main = "PCA, NoDM")
+ vis4 <- plot_responses(res.pca, subnet.id.pca, plot_mode = "pca", datamatrix = D, main = "PCA, DM")
+ 
+ # With original data: VDP overlearns; BIC works; with full covariance data 
+ # With PCA basis: modes detected ok with both VDP and BIC.
+ 
+ # ------------------------------------------------------------------------
+ 
+ # TODO
+ # pca.plot(res, subnet.id)
+ # plot_subnet(res, subnet.id) 
+ }
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
No network provided in function call: assuming fully connected nodes.
convert the network into edge matrix
removing self-links
matching the features between network and datamatrix
Filter the network to only keep the edges with highest mutual information
1 / 5
2 / 5
3 / 5
4 / 5
5 / 5
Compute cost for each variable
Computing model for node 1 / 6
Computing model for node 2 / 6
Computing model for node 3 / 6
Computing model for node 4 / 6
Computing model for node 5 / 6
Computing model for node 6 / 6
independent models done
Computing delta values for edge  1 / 15 

Computing delta values for edge  2 / 15 

Computing delta values for edge  3 / 15 

Computing delta values for edge  4 / 15 

Computing delta values for edge  5 / 15 

Computing delta values for edge  6 / 15 

Computing delta values for edge  7 / 15 

Computing delta values for edge  8 / 15 

Computing delta values for edge  9 / 15 

Computing delta values for edge  10 / 15 

Computing delta values for edge  11 / 15 

Computing delta values for edge  12 / 15 

Computing delta values for edge  13 / 15 

Computing delta values for edge  14 / 15 

Computing delta values for edge  15 / 15 

Combining groups,  6  group(s) left...

Combining groups,  5  group(s) left...

Combining groups,  4  group(s) left...

Combining groups,  3  group(s) left...

No network provided in function call: assuming fully connected nodes.
convert the network into edge matrix
removing self-links
matching the features between network and datamatrix
Filter the network to only keep the edges with highest mutual information
1 / 5
2 / 5
3 / 5
4 / 5
5 / 5
Compute cost for each variable
Computing model for node 1 / 6
Computing model for node 2 / 6
Computing model for node 3 / 6
Computing model for node 4 / 6
Computing model for node 5 / 6
Computing model for node 6 / 6
independent models done
Computing delta values for edge  1 / 15 

Computing delta values for edge  2 / 15 

Computing delta values for edge  3 / 15 

Computing delta values for edge  4 / 15 

Computing delta values for edge  5 / 15 

Computing delta values for edge  6 / 15 

Computing delta values for edge  7 / 15 

Computing delta values for edge  8 / 15 

Computing delta values for edge  9 / 15 

Computing delta values for edge  10 / 15 

Computing delta values for edge  11 / 15 

Computing delta values for edge  12 / 15 

Computing delta values for edge  13 / 15 

Computing delta values for edge  14 / 15 

Computing delta values for edge  15 / 15 

Combining groups,  6  group(s) left...

Combining groups,  5  group(s) left...

Combining groups,  4  group(s) left...

Combining groups,  3  group(s) left...

> 
> proc.time()
   user  system elapsed 
  27.45    0.21   27.65 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: i386-w64-mingw32/i386 (32-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> # 1. vdp.mixt: moodien loytyminen eri dimensiolla, naytemaarilla ja komponenteilla
> #   -> ainakin nopea check
> 
> #######################################################################
> 
> # Generate random data from five Gaussians. 
> # Detect modes with vdp-gm. 
> # Plot data points and detected clusters with variance ellipses
> 
> #######################################################################
> 
> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> #source("~/Rpackages/netresponse/netresponse/R/detect.responses.R")
> #source("~/Rpackages/netresponse/netresponse/R/internals.R")
> #source("~/Rpackages/netresponse/netresponse/R/vdp.mixt.R")
> #dyn.load("/home/tuli/Rpackages/netresponse/netresponse/src/netresponse.so")
> 
> 
> #########  Generate DATA #############################################
> 
> res <- generate.toydata()
> D <- res$data
> component.means <- res$means
> component.sds   <- res$sds
> sample2comp     <- res$sample2comp
> 
> ######################################################################
> 
> # Fit nonparametric Gaussian mixture model
> out <- vdp.mixt(D)
> # out <- vdp.mixt(D, c.max = 3) # try with limited number of components -> OK
> 
> ############################################################
> 
> # Compare input data and results
> 
> ord.out <- order(out$posterior$centroids[,1])
> ord.in <- order(component.means[,1])
> 
> means.out <- out$posterior$centroids[ord.out,]
> means.in <- component.means[ord.in,]
> 
> # Cluster stds and variances
> sds.out <- out$posterior$sds[ord.out,]
> sds.in  <- component.sds[ord.in,]
> vars.out <- sds.out^2
> vars.in <- sds.in^2
> 
> # Check correspondence between input and output
> if (length(means.in) == length(means.out)) {
+    cm <- cor(as.vector(means.in), as.vector(means.out))
+    csd <- cor(as.vector(sds.in), as.vector(sds.out))
+ }
> 
> # Plot results (assuming 2D)
> 
> ran <- range(c(as.vector(means.in - 2*vars.in), 
+                as.vector(means.in + 2*vars.in), 
+ 	       as.vector(means.out + 2*vars.out), 
+ 	       as.vector(means.out - 2*vars.out)))
> 
> plot(D, pch = 20, main = paste("Cor.means:", round(cm,3), "/ Cor.sds:", round(csd,3)), xlim = ran, ylim = ran) 
> for (ci in 1:nrow(means.out))  { add.ellipse(centroid = means.out[ci,], covmat = diag(vars.out[ci,]), col = "red") }
> for (ci in 1:nrow(means.in))  { add.ellipse(centroid = means.in[ci,], covmat = diag(vars.in[ci,]), col = "blue") }
> 
> 
> 
> proc.time()
   user  system elapsed 
   3.23    0.28    3.48 

R version 4.0.3 (2020-10-10) -- "Bunny-Wunnies Freak Out"
Copyright (C) 2020 The R Foundation for Statistical Computing
Platform: x86_64-w64-mingw32/x64 (64-bit)

R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.

R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.

Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.

> 
> # 1. vdp.mixt: moodien loytyminen eri dimensiolla, naytemaarilla ja komponenteilla
> #   -> ainakin nopea check
> 
> #######################################################################
> 
> # Generate random data from five Gaussians. 
> # Detect modes with vdp-gm. 
> # Plot data points and detected clusters with variance ellipses
> 
> #######################################################################
> 
> library(netresponse)
Loading required package: Rgraphviz
Loading required package: graph
Loading required package: BiocGenerics
Loading required package: parallel

Attaching package: 'BiocGenerics'

The following objects are masked from 'package:parallel':

    clusterApply, clusterApplyLB, clusterCall, clusterEvalQ,
    clusterExport, clusterMap, parApply, parCapply, parLapply,
    parLapplyLB, parRapply, parSapply, parSapplyLB

The following objects are masked from 'package:stats':

    IQR, mad, sd, var, xtabs

The following objects are masked from 'package:base':

    Filter, Find, Map, Position, Reduce, anyDuplicated, append,
    as.data.frame, basename, cbind, colnames, dirname, do.call,
    duplicated, eval, evalq, get, grep, grepl, intersect, is.unsorted,
    lapply, mapply, match, mget, order, paste, pmax, pmax.int, pmin,
    pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table,
    tapply, union, unique, unsplit, which, which.max, which.min

Loading required package: grid
Loading required package: minet
Loading required package: mclust
Package 'mclust' version 5.4.6
Type 'citation("mclust")' for citing this R package in publications.
Loading required package: reshape2

netresponse (C) 2008-2020 Leo Lahti et al.

https://github.com/antagomir/netresponse
> #source("~/Rpackages/netresponse/netresponse/R/detect.responses.R")
> #source("~/Rpackages/netresponse/netresponse/R/internals.R")
> #source("~/Rpackages/netresponse/netresponse/R/vdp.mixt.R")
> #dyn.load("/home/tuli/Rpackages/netresponse/netresponse/src/netresponse.so")
> 
> 
> #########  Generate DATA #############################################
> 
> res <- generate.toydata()
> D <- res$data
> component.means <- res$means
> component.sds   <- res$sds
> sample2comp     <- res$sample2comp
> 
> ######################################################################
> 
> # Fit nonparametric Gaussian mixture model
> out <- vdp.mixt(D)
> # out <- vdp.mixt(D, c.max = 3) # try with limited number of components -> OK
> 
> ############################################################
> 
> # Compare input data and results
> 
> ord.out <- order(out$posterior$centroids[,1])
> ord.in <- order(component.means[,1])
> 
> means.out <- out$posterior$centroids[ord.out,]
> means.in <- component.means[ord.in,]
> 
> # Cluster stds and variances
> sds.out <- out$posterior$sds[ord.out,]
> sds.in  <- component.sds[ord.in,]
> vars.out <- sds.out^2
> vars.in <- sds.in^2
> 
> # Check correspondence between input and output
> if (length(means.in) == length(means.out)) {
+    cm <- cor(as.vector(means.in), as.vector(means.out))
+    csd <- cor(as.vector(sds.in), as.vector(sds.out))
+ }
> 
> # Plot results (assuming 2D)
> 
> ran <- range(c(as.vector(means.in - 2*vars.in), 
+                as.vector(means.in + 2*vars.in), 
+ 	       as.vector(means.out + 2*vars.out), 
+ 	       as.vector(means.out - 2*vars.out)))
> 
> plot(D, pch = 20, main = paste("Cor.means:", round(cm,3), "/ Cor.sds:", round(csd,3)), xlim = ran, ylim = ran) 
> for (ci in 1:nrow(means.out))  { add.ellipse(centroid = means.out[ci,], covmat = diag(vars.out[ci,]), col = "red") }
> for (ci in 1:nrow(means.in))  { add.ellipse(centroid = means.in[ci,], covmat = diag(vars.in[ci,]), col = "blue") }
> 
> 
> 
> proc.time()
   user  system elapsed 
   3.48    0.23    3.70