setwd("/home/boussau/Data/Convergenomix/OccigenScaling/toGetForEvaluatingEfficiency2")

Analysis of the moves

d<-read.table("move_stats.tsv", h=T, fill=T, sep="\t")
summary(d)
      proc                                 file                            move       acceptance        mean_time          total_time       
 Min.   : 0.00   diffselbench_1n_24p_200i.out:436   CompMoveFitness(1.0)     : 23   Min.   : 0.1264   Min.   :     3.4   Min.   :2.155e+03  
 1st Qu.: 5.00                                      MoveBaselineFitness(0.3) : 23   1st Qu.:12.2118   1st Qu.: 35934.2   1st Qu.:6.126e+08  
 Median :11.00                                      MoveBaselineFitness(1.0) : 23   Median :30.7426   Median :253273.5   Median :2.109e+09  
 Mean   :11.39                                      MoveFitnessShifts(1, 0.3): 23   Mean   :42.7595   Mean   :227022.0   Mean   :2.953e+09  
 3rd Qu.:17.00                                      MoveFitnessShifts(1, 1)  : 23   3rd Qu.:80.3671   3rd Qu.:345168.0   3rd Qu.:4.694e+09  
 Max.   :23.00                                      MoveMasks(weight)        : 23   Max.   :94.6349   Max.   :674842.0   Max.   :8.928e+09  
                                                    (Other)                  :298   NA's   :46                                              
library(plyr)

Averages

library(plyr)
ddply(d,~move,summarise,success=mean(acceptance),mean_time=mean(mean_time), total_time=mean(total_time))

Maximums

ddply(d,~move,summarise,success=max(acceptance),mean_time=max(mean_time), total_time=max(total_time))

Heterogeneity of computing time across processes

times_per_proc <- ddply(d,~proc,summarise, total_time=sum(total_time))
times_per_proc
plot(times_per_proc$total_time, ylab="Time per processor", pch=20, cex=2)

Analysis of the traces

d1 <- read.table("diffselbench_50_bboussau_1n_24p_200i.trace", h=T, comment.char = "/")
#d4 <- read.table("diffselbench_50_bboussau_4n_96p_200i.trace", h=T, comment.char = "/")
#d16 <- read.table("diffselbench_50_bboussau_16n_384p_200i.trace", h=T, comment.char = "/")
#d64 <- read.table("diffselbench_50_bboussau_64n_1536p_200i.trace", h=T, comment.char = "/")
plot(d1$X.logprior, t="l")

library(coda)
md1 <- mcmc(d1)
print(summary(md1))

Iterations = 1:63
Thinning interval = 1 
Number of chains = 1 
Sample size per chain = 63 

1. Empirical mean and standard deviation for each variable,
   plus standard error of the mean:

                    Mean        SD  Naive SE Time-series SE
X.logprior    -1.325e+06 6.422e+04 8.092e+03      2.218e+04
lnL           -6.960e+06 4.722e+04 5.949e+03      5.949e+03
length         1.110e+01 2.843e-01 3.581e-02      3.581e-02
invshape       2.680e-01 1.528e-02 1.926e-03      3.182e-03
meanwidth      3.760e+00 2.966e-01 3.737e-02      9.998e-02
meaneps        7.314e-03 9.560e-04 1.204e-04      1.204e-04
pi1            3.994e-01 7.376e-02 9.293e-03      3.790e-02
mean           1.727e-01 5.768e-02 7.266e-03      2.481e-02
invconc        1.504e-01 5.093e-02 6.417e-03      2.100e-02
nucstatcenter  1.356e+00 3.742e-03 4.714e-04      6.138e-04
invconc.1      2.027e-02 1.783e-03 2.246e-04      3.124e-04
relratecenter  1.604e+00 1.975e-02 2.488e-03      3.765e-03
invconc.2      5.281e-03 1.401e-03 1.766e-04      1.766e-04

2. Quantiles for each variable:

                    2.5%        25%        50%        75%      97.5%
X.logprior    -1.464e+06 -1.332e+06 -1.300e+06 -1.289e+06 -1.283e+06
lnL           -6.982e+06 -6.955e+06 -6.954e+06 -6.953e+06 -6.951e+06
length         1.087e+01  1.104e+01  1.107e+01  1.112e+01  1.118e+01
invshape       2.534e-01  2.629e-01  2.659e-01  2.726e-01  2.907e-01
meanwidth      3.587e+00  3.601e+00  3.647e+00  3.772e+00  4.457e+00
meaneps        6.969e-03  7.132e-03  7.185e-03  7.280e-03  7.408e-03
pi1            2.836e-01  3.468e-01  3.779e-01  4.796e-01  4.990e-01
mean           5.994e-02  1.297e-01  1.816e-01  2.130e-01  2.617e-01
invconc        1.002e-02  1.385e-01  1.666e-01  1.862e-01  1.972e-01
nucstatcenter  1.350e+00  1.354e+00  1.356e+00  1.358e+00  1.362e+00
invconc.1      1.640e-02  1.958e-02  2.040e-02  2.121e-02  2.330e-02
relratecenter  1.595e+00  1.599e+00  1.601e+00  1.602e+00  1.637e+00
invconc.2      4.541e-03  4.933e-03  5.073e-03  5.281e-03  6.253e-03
a<-autocorr.plot(d1, auto.layout = F)

summary(d1)
   X.logprior            lnL               length         invshape        meanwidth        meaneps              pi1              mean        
 Min.   :-1694670   Min.   :-7325820   Min.   :10.67   Min.   :0.1748   Min.   :3.587   Min.   :0.006923   Min.   :0.2589   Min.   :0.05149  
 1st Qu.:-1331735   1st Qu.:-6954505   1st Qu.:11.04   1st Qu.:0.2629   1st Qu.:3.601   1st Qu.:0.007132   1st Qu.:0.3468   1st Qu.:0.12973  
 Median :-1300360   Median :-6953590   Median :11.07   Median :0.2659   Median :3.647   Median :0.007185   Median :0.3779   Median :0.18157  
 Mean   :-1324538   Mean   :-6960312   Mean   :11.10   Mean   :0.2680   Mean   :3.760   Mean   :0.007314   Mean   :0.3994   Mean   :0.17274  
 3rd Qu.:-1289050   3rd Qu.:-6952630   3rd Qu.:11.12   3rd Qu.:0.2726   3rd Qu.:3.772   3rd Qu.:0.007280   3rd Qu.:0.4796   3rd Qu.:0.21297  
 Max.   :-1280860   Max.   :-6950730   Max.   :13.22   Max.   :0.2928   Max.   :5.479   Max.   :0.014734   Max.   :0.4998   Max.   :0.27461  
    invconc         nucstatcenter     invconc.1       relratecenter     invconc.2       
 Min.   :0.003104   Min.   :1.350   Min.   :0.01289   Min.   :1.595   Min.   :0.004428  
 1st Qu.:0.138519   1st Qu.:1.354   1st Qu.:0.01958   1st Qu.:1.599   1st Qu.:0.004933  
 Median :0.166577   Median :1.356   Median :0.02040   Median :1.601   Median :0.005073  
 Mean   :0.150441   Mean   :1.356   Mean   :0.02027   Mean   :1.604   Mean   :0.005281  
 3rd Qu.:0.186205   3rd Qu.:1.358   3rd Qu.:0.02121   3rd Qu.:1.602   3rd Qu.:0.005281  
 Max.   :0.198884   Max.   :1.374   Max.   :0.02428   Max.   :1.745   Max.   :0.015857  
d1ess <- apply(d1, 2, effectiveSize)
plot( d1ess, xaxt="n", main= "ESS of run with 1 node", xlab="", ylab = "ESS")
#axis(1, at=, labels=names(d1ess))
axis(1, at=1:length(d1ess), labels = FALSE)
text(1:length(d1ess), par("usr")[3] - 8.9, labels = names(d1ess), srt = 90, pos = 1, xpd = TRUE)

Percent of time spent in communications + waiting

100*(8.28614e+07 - 544039 - 6.34893e+07 - 7.15199e+06) / 8.28614e+07

[1] 14.09109

We spend 14% of the time in communications or waiting.

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