Network Latent Source Separation (NLSS), a blind source separation algorithm designed for network data.
Install the released version of nlss from Github with:
devtools::install_github("benwu233/nlss")This is a basic example shows the implement of nlss package:
Load the package and generate three true latent source networks.
library(nlss)
set.seed(716)
signal = gen_sources(100)Draw the true sources.
library(gplots)
heatmap.net(signal$S,lim=c(0,1),
community = signal$community,
color = colorpanel(n = 100, low = "white", high = "black"))
Generate network data with NLSS.
sim0 = sim_NLSS(n = 150, alpha_0 = c(0.8,0.8,0.8),
alpha_1 = 0.8, sd0 = 0.25, signal$S)Draw samples of the network data.
heatmap.net(sim0$Xc[1:3,],lim=c(0,1),
community = signal$community,
color = colorpanel(n = 100, low = "white", high = "black"))
Solve with the MCMC algorithm:
data0 = thr_byrow(sim0$Xc,0.1)
res = NLSS(data=data0, q=3, states = c(0,1), state0 = 0,
total_iter = 2000, burn_in = 0, thin = 10,
show_step=1000, joint=TRUE,
q0 = 3)Summarize the results:
sum_res = NLSS_sum(res,th=0.95, nstart = 1, nend = 100)Print the estimated source networks:
heatmap.net(sum_res$S,lim=c(0,1),
community = signal$community,
color = colorpanel(n = 100, low = "white", high = "black"))
Match and compare with the true sources:
S_match = match_source(signal$S, sum_res$S)
#> | | | 0% | |======== | 11% | |================ | 22% | |======================= | 33% | |=============================== | 44% | |======================================= | 56% | |=============================================== | 67% | |====================================================== | 78% | |============================================================== | 89% | |======================================================================| 100%
lapply(1:3, function(i){caret::confusionMatrix(as.factor(S_match[i,]), as.factor(signal$S[i,]), positive = "1" )} )
#> [[1]]
#> Confusion Matrix and Statistics
#>
#> Reference
#> Prediction 0 1
#> 0 4250 6
#> 1 0 694
#>
#> Accuracy : 0.9988
#> 95% CI : (0.9974, 0.9996)
#> No Information Rate : 0.8586
#> P-Value [Acc > NIR] : < 2e-16
#>
#> Kappa : 0.995
#>
#> Mcnemar's Test P-Value : 0.04123
#>
#> Sensitivity : 0.9914
#> Specificity : 1.0000
#> Pos Pred Value : 1.0000
#> Neg Pred Value : 0.9986
#> Prevalence : 0.1414
#> Detection Rate : 0.1402
#> Detection Prevalence : 0.1402
#> Balanced Accuracy : 0.9957
#>
#> 'Positive' Class : 1
#>
#>
#> [[2]]
#> Confusion Matrix and Statistics
#>
#> Reference
#> Prediction 0 1
#> 0 3570 84
#> 1 0 1296
#>
#> Accuracy : 0.983
#> 95% CI : (0.979, 0.9864)
#> No Information Rate : 0.7212
#> P-Value [Acc > NIR] : < 2.2e-16
#>
#> Kappa : 0.957
#>
#> Mcnemar's Test P-Value : < 2.2e-16
#>
#> Sensitivity : 0.9391
#> Specificity : 1.0000
#> Pos Pred Value : 1.0000
#> Neg Pred Value : 0.9770
#> Prevalence : 0.2788
#> Detection Rate : 0.2618
#> Detection Prevalence : 0.2618
#> Balanced Accuracy : 0.9696
#>
#> 'Positive' Class : 1
#>
#>
#> [[3]]
#> Confusion Matrix and Statistics
#>
#> Reference
#> Prediction 0 1
#> 0 4465 6
#> 1 0 479
#>
#> Accuracy : 0.9988
#> 95% CI : (0.9974, 0.9996)
#> No Information Rate : 0.902
#> P-Value [Acc > NIR] : < 2e-16
#>
#> Kappa : 0.9931
#>
#> Mcnemar's Test P-Value : 0.04123
#>
#> Sensitivity : 0.98763
#> Specificity : 1.00000
#> Pos Pred Value : 1.00000
#> Neg Pred Value : 0.99866
#> Prevalence : 0.09798
#> Detection Rate : 0.09677
#> Detection Prevalence : 0.09677
#> Balanced Accuracy : 0.99381
#>
#> 'Positive' Class : 1
#>