-
Notifications
You must be signed in to change notification settings - Fork 0
/
learning-mix.R
178 lines (123 loc) · 4.38 KB
/
learning-mix.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
library(mixdist)
?mix
rm(list=ls())
data(pike65)
pike65
?plot
plot(pike65, type="s")
data(pikepar)
pikepar
fitpike1 = mix(pike65, pikepar, "lnorm", constr=mixconstr(consigma="CCV"), emsteps=3)
names(fitpike1)
plot(fitpike1) #wow!!!
fitted(fitpike1)
data(pike65sg)
pike65sg
fitpike2 = mix(pike65sg, pikepar, "lnorm", emsteps=3, usecondit=TRUE)
fitpike2
plot(fitpike2)
data(bindat)
bindat
data(binpar)
binpar
fitbin1 = mix(bindat, binpar, "binom", constr=mixconstr(consigma="BINOM", size=c(20,20,20,20)))
plot(fitbin1)
fitbin2 = mix(bindat, binpar, "binom", constr=mixconstr(conpi="PFX", fixpi=c(TRUE,TRUE,TRUE,TRUE), consigma="BINOM", size=c(20,20,20,20)))
plot(fitbin2)
# Looking at my data
load("~/Desktop/ops.RData")
h1=hist(op1$latency_ms)
names(h1)
h1$breaks # left side of break
h3=hist(op3$latency_ms, breaks=100)
op3GroupedData = matrix(c(h3$breaks, h3$counts, 0), nrow=length(h3$breaks), ncol=2)
plot(op3GroupedData, type="s")
length(h3$breaks)
length(h3$counts)
# using mixgroup
data(pikeraw)
pikeraw
colnames(pikeraw)
mixgroup(pikeraw)
pikemd = mixgroup(pikeraw, breaks=c(0,seq(19.75, 65.75, 2), 80))
plot(pikemd)
pikemd = mixgroup(pikeraw, breaks=c(0,seq(19.75, 65.75, 2), 80), usecondit=TRUE, k=5)
plot(pikemd)
mixgroup(pikeraw, usecondit=TRUE)
mixgroup(pikeraw, usecondit=TRUE, k=3)
mixgroup(pikeraw, usecondit=TRUE, k=8)
# trying mixgroup on my data
op1Grouped=mixgroup(op1$latency_ms)
quantile(op1$latency_ms, 0.99)
quantile(op1$latency_ms, 0.999)
quantile(op1$latency_ms, 0.9999)
# might want to try pred at 3 9's
# using mixparam
mixparam(mu=c(20,30,40), sigma=c(2,3,4))
mixparam(c(20,30,40), c(3), c(0.15, 0.78, 0.07))
# try using mixparam for my data
library(mclust)
source("/Users/radlab/Desktop/ksauer/Desktop/scads/experiments/client/performance/logparsing/src/main/R/emFitAndSample.R")
op=op3
k=3
?mixgroup
opGrouped=mixgroup(op$latency_ms, breaks=100)
rdmInit = randomInitEM(length(op$latency_ms), k)
em = mstep(modelName="V", data=op$latency_ms, z=rdmInit)
params = em$parameters
# sort so means are in ascending order;
# use same order for sd, pi
sortedMean = sort(params$mean, index.return=TRUE)
paramsMeanSorted = matrix(c(sortedMean$x, sqrt(params$variance$sigmasq[sortedMean$ix]), params$pro[sortedMean$ix]), nrow=length(sortedMean$x), ncol=3)
colnames(paramsMeanSorted)=c("mean", "sd", "pi")
mixParam = mixparam(mu=paramsMeanSorted[,"mean"], sigma=paramsMeanSorted[,"sd"], pi=paramsMeanSorted[,"pi"])
mixParamA = mixparam(mu=paramsMeanSorted[,"mean"], sigma=paramsMeanSorted[,"sd"])
# try it out!
fit = mix(opGrouped, mixParam, dist="gamma", print.level=2, constr=mixconstr(consigma="CCV"), emsteps=0)
plot(fit)
# this works pretty well!
#mixParam=mixparam(mu=c(3, 17, 28), sigma=c(1,1,1))
mixParam=mixparam(mu=c(3, 25), sigma=c(2,12.5))
fit = mix(opGrouped, mixParam, dist="gamma", print.level=2, constr=mixconstr(consigma="SFX", fixsigma=c(TRUE,TRUE)), emsteps=0)
#fit = mix(opGrouped, mixParam, dist="gamma", print.level=2, emsteps=0)
plot(fit)
# this doesn't work yet
mixParam=mixparam(mu=c(3, 25, 26), sigma=c(1,1,1))
fit = mix(opGrouped, mixParam, dist="weibull", print.level=2, constr=mixconstr(consigma="CCV"), emsteps=0)
plot(fit)
names(fit)
fit$parameters$mu
# try sampling from the fitted distr
sampleFromGammaMixtureModel = function(params, numSamples) {
numMixtures = length(params$mu)
# Get scale, shape params for each component
scale = vector(mode="numeric", length=numMixtures)
shape = vector(mode="numeric", length=numMixtures)
for (i in 1:numMixtures) {
shape = (params$mu[i]^2)/(params$sigma[i]^2)
scale = params$mu[i]/shape
scale[i] = scale
shape[i] = shape
}
# Get samples
mixture = vector(mode="numeric", length=numSamples)
samples = vector(mode="numeric", length=numSamples)
for (i in 1:numSamples) {
# Get mixture component
mixture[i] = sample(x=seq(from=1,to=numMixtures,by=1), size=1, prob=params$pi)
# Get sample from that mixture's gamma
samples[i] = rgamma(n=1, shape=shape[mixture[i]], scale=scale[mixture[i]])
}
return(samples)
#return(mixture)
}
gammaSamples = sampleFromGammaMixtureModel(fit$parameters, 1000)
length(which(gammaSamples == 1))/length(gammaSamples)
length(which(gammaSamples == 2))/length(gammaSamples)
par(mfrow=c(2,1))
xmax=25+quantile(op$latency_ms, 0.99)
hist(op$latency_ms, xlim=c(0,xmax), breaks=100)
hist(gammaSamples, xlim=c(0,xmax), breaks=100)
par(mfrow=c(2,1))
plot(fit)
hist(op$latency_ms, breaks=100)