-
Notifications
You must be signed in to change notification settings - Fork 0
/
AnylVarImp.Rmd
338 lines (302 loc) · 18 KB
/
AnylVarImp.Rmd
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
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
---
title: "Variable importance"
---
is the signal clean but containing rare usefullness or noisy and always present?
load all varimp files and check problems and remove some problems
```{r}
col_typed <- "cccddccccccddddddcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcdcd?????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????"
#71 per row so 142+ more than 20?
#167
try({setwd(mainDir)})
mainDir<-getwd()
subDir<-"QSlinks_simple"
#dir.create(file.path(mainDir, subDir))
require(readr)
require(netCoin)
require(data.table)
require(naniar)
require(VIM)
require(ggplot2)
require(stringr)
resultDir<-"ACEREBOUT"
#setwd(resultDir)
setwd(file.path(mainDir, resultDir))
QSlink3ACE <- read_csv("importanceQSlink3ACEREBOUTwindowsx64.csv",col_types = col_typed)
resetACE <- read_csv("test after which reset.csv")
names(resetACE)<-c("date","model","task","fold","cv","arch","arch2","pc")
#QSlinkImp3ACE <- read_csv("importanceQSlink3ACEREBOUTwindowsx64.csv",col_types = col_typed)
#summary(QSlink3ACE[,1:30])
#spec(QSlink3ACE)
prACE<-problems(QSlink3ACE)
unique(prACE$actual)
#prACE[!str_detect(prACE$actual,"columns"),]
rowrem<-prACE[!str_detect(prACE$actual,"columns"),]$row
QSlink3ACE<-QSlink3ACE[-rowrem,]
setwd(file.path(mainDir))
resultDir<-"HOPPER"
#setwd(resultDir)
setwd(file.path(mainDir, resultDir))
QSlink3HOP <- read_csv("importanceQSlink3HOPPERwindowsx64.csv",col_types = col_typed)
resetHOP <- read_csv("test after which reset.csv")
names(resetHOP)<-c("date","model","task","fold","cv","arch","arch2","pc")
#summary(QSlink3ACE[,1:30])
#spec(QSlink3HOP)
prHOP<-problems(QSlink3HOP)
unique(prHOP$actual)
#prHOP[!str_detect(prHOP$actual,"columns"),]
rowrem<-prHOP[!str_detect(prHOP$actual,"columns"),]$row
QSlink3HOP<-QSlink3HOP[-rowrem,]
setwd(file.path(mainDir))
#names(QSlink3HOP)
resultDir<-"LAPTOPBBQ"
#setwd(resultDir)
setwd(file.path(mainDir, resultDir))
QSlink3BBQ <- read_csv("importanceQSlink3LAPTOPBBQwindowsx64.csv",col_types = col_typed)
resetBBQ <- read_csv("test after which reset.csv")
names(resetBBQ)<-c("date","model","task","fold","cv","arch","arch2","pc")
#summary(QSlink3ACE[,1:30])
#spec(QSlink3BBQ)
prBBQ<-problems(QSlink3BBQ)
unique(prBBQ$actual)
#prBBQ[!str_detect(prBBQ$actual,"columns"),]
rowrem<-prBBQ[!str_detect(prBBQ$actual,"columns"),]$row
QSlink3BBQ<-QSlink3BBQ[-rowrem,]
setwd(file.path(mainDir))
```
tibble into data.table
```{r}
require(data.table)
DF<-as.data.table(QSlink3ACE)
DF1<-as.data.table(QSlink3HOP)
DF2<-as.data.table(QSlink3BBQ)
DF<-rbindlist(list(DF,DF1,DF2),fill = T)
rm(DF1,DF2,QSlink3BBQ,QSlink3HOP,QSlink3ACE)
reset<-as.data.table(resetACE)
reset1<-as.data.table(resetHOP)
reset2<-as.data.table(resetBBQ)
reset<-rbindlist(list(reset,reset1,reset2),fill = T,idcol=T)
rm(reset1,reset2,resetACE,resetHOP,resetBBQ)
gc()
```
check rename remove nas
```{r}
print("any fails in 4th and fifith coulums? success and transtarg")
sum(str_detect(DF$X4,"FAIL"),na.rm = T)
sum(str_detect(DF$X5,"FAIL"),na.rm = T)
#str(DF)
head(DF)
unique(DF$X13)
DFnam<-c("modelset","model","datetime","success","transTarg","task","missing","append","transform","pc","expirament","fold","maxfold","seed","seedit","foldseed","RMSEmean")
colNum<-T
firCol<-length(DFnam)+1
for (i in (firCol):dim(DF)[2]) {
if(colNum){
DFnam<-c(DFnam,paste0("Col",i))
colNum<-F
} else {
DFnam<-c(DFnam,paste0("Val",i))
colNum<-T
}
}
names(DF)<-DFnam
summary((DF$Val27))
DF[(is.na(Val29) & is.na(Val27)),]
DF[(is.infinite(Val29)),]
DF<-DF[!(is.na(Val29) & is.na(Val27)),]
DF<-DF[!(is.infinite(Val29)),]
print("summary of first and 4th and 5th valuse")
summary((DF$Val19));summary((DF$Val29));summary((DF$Val27))
```
select best from pessimist file ! not finished just copied from outside
```{r}
load(file="routuDFpessimistUnique.RData")
if(F){
(bestest<-Unq[(1)])
nnd<-which(names(Unq)=="datePOSIX")
strt<-which(names(Unq)=="btp8")
Unqn<-Unq[(is.na(btp11))]
Unqf<-Unq[(!is.na(btp11))]
Unqf[,dum1:=0]
Unqf[,dum2:=0]
Unqf[,dum3:=0]
Unqn[,dum1:=0]
Unqn[,dum2:=0]
Unqn[,dum3:=0]
colord<-names(Unqf)
ol<-length(colord)
(colord<-c(colord[1:strt],colord[(ol-2):ol],colord[(strt+1):(nnd-4)],colord[(nnd):(ol-3)]))
setcolorder(Unqf,colord)
}
```
before melting casting massaging w/e must separate by subject?
otherwise calculation time and memory is huge
poor organization; separated and duplicated results per each recipe into tiddy table
FUNCTION to convert to sparse matrix
```{r}
chosenToSparse<-function(datFr)
{
cols<-which(str_detect(names(datFr),fixed("Col")))
vals<-which(str_detect(names(datFr),fixed("Val")))
(fir<-min(cols))
(endd<-max(cols))
(lst<-max(cols))
if(lst==dim(datFr)[2]) {
datFr[,(lst):=NULL]
cols<-cols[-length(cols)]
lst<-lst-2
}#dim(datFr)
strt<-fir#names(datFr)[strt]
nnd<-lst+1
prd<-data.table(allColNm=unique(as.vector(datFr[,(cols),with=F][[1]])))
length(unique(datFr$keyVarImp))
i<-unique(datFr$keyVarImp)[10]
for(i in unique(datFr$keyVarImp)){
wrk<-datFr[(keyVarImp==i)][,(strt:nnd),with=F]
wrk<-unique(wrk)
indx<-(unlist(wrk[,seq(by=2,from=1,to=dim(wrk)[2]),with=F]))
val<-(unlist(wrk[,seq(by=2,from=2,to=dim(wrk)[2]),with=F]))
unique(indx)
#unique(val)
wrk<-data.table(indx,val)
wrk$val<-as.numeric(wrk$val)
dim(wrk)
wrk<-wrk[(order(indx))]
#wrk[,sdw:=sd(val),by=indx]
#wrk[,Nw:=.N,by=indx]
wrk[,val:=mean(val),by=indx]
wrk<-unique(wrk)
if(dim(wrk)[1]!=length(unique(wrk$indx))) stop("more work, multiple valuse for same index")
names(wrk)<-c("indx",paste0(i,"val"))#,paste0(i,"sd"),paste0(i,"N"))
prd<-merge(prd, wrk, by.x = "allColNm", by.y = "indx", all = T)
}
return(prd)
}
rm(datFr)
```
select by task, run choice to sparse,
```{r}
require(stringr)
Unq[,keyVarUnq:=paste0( algomodel, transTarg, task, transform, expirament)]
DF[,keyVarImp:=paste0(modelset, model, transTarg, task, transform, expirament)]
DF[,keyVarUnq:=paste0( model, transTarg, task, transform, expirament)]
(tsks<-unique(Unq$task))
#unique(DF$task)
its<- "QSlinks03_95"#"QSlinks115_60"#tsks[1]
DFs<-DF[(task == its )]
DFs<-DFs[(str_detect(modelset,fixed("hold")) )]
prds<-chosenToSparse(DFs[(keyVarUnq %in% Unq$keyVarUnq)])
#datFr<-DF[(task == its )]
dim(prds)
```
clean, outliers, features
```{r}
#implemented outlier removal by hand because didn't bother to lookup sparse matrix
lic<-dim(prds)[2]
IQn<-quantile(unlist(prds[,2:lic,with=F]),c(.01,.99),na.rm = T)
IQr<-(IQn[2]-IQn[1])*20
IQn[1]<-IQn[1]-IQr
IQn[2]<-IQn[2]+IQr
if((IQn[2]-IQn[1] > 10)){
wiOut<-names(prds)[which(apply(prds[,2:lic,with=F],2,max,na.rm=T)>IQn[2] | apply(prds[,2:lic,with=F],2,min,na.rm=T)< IQn[1])+1]
i<-wiOut[1]
for(i in wiOut){
toobig<-as.vector(prds[,i,with=F]>IQn[2])
toobig[is.na(toobig)]<-F
prds[(toobig),(i):=IQn[2]]
toobig<-as.vector(prds[,i,with=F]<IQn[1])
toobig[is.na(toobig)]<-F
prds[(toobig),(i):=IQn[1]]
}
}
nonna<-function(x){
return(sum(!is.na(x)))
}
prds$countVar<-apply(prds[,2:lic,with=F],1,nonna)
prds$minVar<-apply(prds[,2:lic,with=F],1,min,na.rm=T)
prds$maxVar<-apply(prds[,2:lic,with=F],1,max,na.rm=T)
prds$meanVar<-apply(prds[,2:lic,with=F],1,mean,na.rm=T)
```
plots to understand Varimp
```{r}
plot(prds$countVar, prds$maxVar)
plot(prds$countVar, prds$meanVar)
hist(prds$countVar)
hist(prds$meanVar)
plot(prds$minVar, prds$maxVar)
plot(prds$meanVar, prds$maxVar)
#prds[(order(meanVar))]$allColNm
```
at this point about to give up because converting column names would be hard and code to fix exists but lets see it anyway
```{r}
actNamesQSlinks03_95 <- c("page","searxinstance","tries","success","time","isPDF","enginebing","engineduckduckgo","enginefaroo","engineggl","enginegoogle","enginegoogle_scholar","enginemojeek","engineyacy","engineyahoo","string.connectordb.","string.markwk.","string.open.humans.","string.open.mhealth.","string.personal_dashboard.","stringBiofourmis","stringBodyTrack","stringconnectordb.io","stringexist.io","stringFitnessSyncer.fluxtream","stringFitnessSyncer.HealthKit","stringFitnessSyncer.zenobase","stringfluxtream.zenobase","stringHealthKit","stringHealthKit.BodyTrack.exist.io","stringHealthKit.BodyTrack.humanapi.co","stringHealthKit.BodyTrack.open.humans"," "," "," "," "," "," ","stringHealthKit.fluxtream.zenobase","stringHealthKit.humanapi.co.betterself.io","stringHealthKit.humanapi.co.connectordb.io","stringHealthKit.humanapi.co.gyrosco.pe","stringHealthKit.humanapi.co.open.humans","stringHealthKit.humanapi.co.open.mhealth","stringHealthKit.humanapi.co.openhumans.org","stringHealthKit.humanapi.co.openmhealth.org","stringHealthKit.tapiriik.connectordb.io","stringHealthKit.tapiriik.gyrosco.pe","stringHealthKit.tapiriik.open.mhealth","stringHealthKit.tapiriik.zenobase","stringHealthKit.tictrac.exist.io","stringHealthKit.tictrac.fluxtream","stringHealthKit.tictrac.gyrosco.pe","stringHealthKit.tictrac.humanapi.co","stringHealthKit.tictrac.open.humans","stringHealthKit.tictrac.open.mhealth","stringHealthKit.tictrac.zenobase","stringHealthKit.validic.BodyTrack","stringHealthKit.validic.exist.io","stringHealthKit.validic.fluxtream","stringHealthKit.validic.gyrosco.pe","stringHealthKit.validic.humanapi.co","stringHealthKit.validic.open.humans","stringHealthKit.validic.open.mhealth","stringHealthKit.validic.openmhealth.org","stringHealthKit.validic.tictrac","stringHealthKit.validic.zenobase","stringHealthKit.zenobase.betterself.io","stringHealthKit.zenobase.connectordb.io","stringHealthKit.zenobase.gyrosco.pe","stringHealthKit.zenobase.open.humans","stringHealthKit.zenobase.open.mhealth","stringHealthKit.zenobase.openmhealth.org","stringhealthvault","stringopen.mhealth","stringopenhumans.org","stringopenmhealth.org","stringpersonal_dashboard","stringqs_ledger","stringtictrac","stringtsubery.quantifier","stringvalidic","sCntQstr","sCntEngine","sRareSite","sRareQstr","sRareEngine","s.score","sRank","sStrongSubQstr","searchCount","sSearchBrevity","sRareSiteBySearch","sSimpleBySearch","sLinProBySearch","sDnchar","sDiNA","fluxtream","healthvault","tapiriik","tictrac","humanapi.co","zenobase","exist.io","quantimodo","openmhealth.org","betterself.io","connectordb.io","connectordb","qs_ledger","tsubery","quantifiedmind","habitica","vitadock","openhumans","humanapi","openmhealth","sDCountWords","sDUWfract","sDUWlog","sDUWsqrt","pageMin","pageMax","searxinstanceMax","triesMax","successMax","timeMax","enginebingMax","engineduckduckgoMax","enginegglMin","enginegglMax","enginegoogleMax","enginegoogle_scholarMax","enginemojeekMax","engineqwantMax","engineredditMax","engineseznamMax","enginestartpageMax","engineyacyMax","engineyahooMax","string.connectordb.Min","string.connectordb.Max","string.markwk.Max","string.open.humans.Min","string.open.humans.Max","string.open.mhealth.Max","string.personal_dashboard.Max","stringBodyTrackMax","stringconnectordb.ioMin","stringexist.ioMax","stringFitnessSyncerMax","stringfluxtreamMin","stringfluxtreamMax","stringfluxtream.zenobaseMax","stringHealthKitMin","stringHealthKit.BodyTrack.connectordb.ioMax","stringHealthKit.BodyTrack.exist.ioMax","stringHealthKit.BodyTrack.humanapi.coMax","stringHealthKit.BodyTrack.open.humansMax"," "," "," "," stringHealthKit.BodyTrack.zenobaseMax","stringHealthKit.fluxtream.exist.ioMax","stringHealthKit.fluxtream.open.humansMax","stringHealthKit.fluxtream.open.mhealthMax","stringHealthKit.fluxtream.openhumans.orgMax","stringHealthKit.fluxtream.openmhealth.orgMax","stringHealthKit.fluxtream.zenobaseMax","stringHealthKit.healthvaultMax","stringHealthKit.humanapi.co.gyrosco.peMax","stringHealthKit.humanapi.co.open.mhealthMax","stringHealthKit.humanapi.co.openhumans.orgMax")#
otherhalf <- c("stringHealthKitlthMin","stringHeargMax","stringHealthKit.humanapitsubery.quantifierMax","stringHealthKit.tapiriik.gyrosco.peMax","stringHealthKit.tapiriik.open.mhealthMax","stringHealthKit.tapiriik.zenobaseMax","stringHealthKit.tictrac.exist.ioMax","stringHealthKit.tictrac.fluxtreamMax","stringHealthKit.tictrac.gyrosco.peMax","stringHealthKit.tictrac.humanapi.coMax","stringHealthKit.tictrac.openhumans.orgMax","stringHealthKit.tictrac.openmhealth.orgMax","stringHealthKit.tictrac.zenobaseMax","stringHealthKit.validic.BodyTrackMax","stringHealthKit.validic.connectordb.ioMax","stringHealthKit.validic.exist.ioMax","stringHealthKit.validic.fluxtreamMax","stringHealthKit.validic.gyrosco.peMax","stringHealthKit.validic.humanapi.coMax","stringHealthKit.validic.markwkMax","stringHealthKit.validic.open.humansMax","stringHealthKit.validic.open.mhealthMax","stringHealthKit.validic.openhumans.orgMax","stringHealthKit.validic.openmhealth.orgMax","stringHealthKit.validic.tictracMax","stringHealthKit.validic.zenobaseMax","stringHealthKit.zenobase.betterself.ioMax","stringHealthKit.zenobase.gyrosco.peMin","stringHealthKit.zenobase.gyrosco.peMax","stringHealthKit.zenobase.markwkMax","stringHealthKit.zenobase.open.humansMax","stringHealthKit.zenobase.openmhealth.orgMax","stringHealthKit.zenobase.quantimodoMax","stringHealthKit.zenobase.tsubery.quantifierMin","stringhealthvaultMax","stringhuman.apiMax","stringmyfitnesspalMax","stringopen.mhealthMax","stringopenhumans.orgMax","stringopenmhealth.orgMin","stringqs_ledgerMax","stringquantimodoMin","stringvalidicMax","stringzenobaseMax","sRareQstrMin","sRareQstrMax","sRareEngineMax","sRankMin","searchCountMin","searchCountMax","sSearchBrevityMin","sRareSiteBySearchMin","sRareSiteBySearchMax","sSimpleBySearchMin","sSimpleBySearchMax","sLinProBySearchMin","sLinProBySearchMax")
#
#print(actNamesQSlinks03_95)
#print(otherhalf)
actNamesQSlinks03_95<-c(actNamesQSlinks03_95,otherhalf)
```
```{r}
actNamesQSlinks115_60<-c("page","tries","isPDF","engineduckduckgo","enginegoogle","engineyahoo","stringBiofourmis","stringFitnessSyncer.HealthKit","stringHealthKit.BodyTrack.exist.io","stringHealthKit.BodyTrack.humanapi.co","stringHealthKit.fluxtream.open.humans","stringHealthKit.fluxtream.open.mhealth","stringHealthKit.humanapi.co.connectordb.io","stringHealthKit.humanapi.co.open.mhealth","stringHealthKit.humanapi.co.openmhealth.org","stringHealthKit.tictrac.fluxtream","stringHealthKit.tictrac.humanapi.co","stringHealthKit.tictrac.open.mhealth","stringHealthKit.validic.BodyTrack","stringHealthKit.validic.exist.io","stringHealthKit.validic.gyrosco.pe","stringHealthKit.validic.humanapi.co","stringHealthKit.validic.open.mhealth","stringHealthKit.validic.openmhealth.org","stringHealthKit.zenobase.betterself.io","stringHealthKit.zenobase.gyrosco.pe","stringHealthKit.zenobase.open.mhealth","sRareSite","sStrongSubQstr","sRareSiteBySearch","sDiNA","healthvault","humanapi.co","vitadock","open mhealth","sDUWfract","pageMin","searxinstanceMax","enginegglMin","enginegoogle_scholarMax","engineyacyMax","stringBodyTrackMax","stringexist.ioMax","stringfluxtream.zenobaseMax","stringHealthKit.BodyTrack.openmhealth.orgMax","stringHealthKit.fluxtream.openmhealth.orgMax","stringHealthKit.humanapi.co.openhumans.orgMax","stringHealthKit.tapiriik.gyrosco.peMax","stringHealthKit.tapiriik.open.mhealthMax","stringHealthKit.tictrac.exist.ioMax","stringHealthKit.tictrac.gyrosco.peMax","stringHealthKit.tictrac.openmhealth.orgMax","stringHealthKit.tictrac.zenobaseMax","stringHealthKit.validic.open.humansMax","stringHealthKit.zenobase.connectordb.ioMax","stringHealthKit.zenobase.gyrosco.peMax","stringHealthKit.zenobase.openmhealth.orgMax","stringzenobaseMax","sRareQstrMax","sSimpleBySearchMin")
for(i in length(actNamesQSlinks03_95):1){
(prds$allColNm<-str_replace(prds$allColNm,paste0("V",i),actNamesQSlinks03_95[i]))
}
prds[(order(meanVar))]$allColNm[1:30]
prds[(order(maxVar))]$allColNm[1:30]
prds[(order(countVar))]$allColNm[1:30]
```
alt mean that equates all ignored with 1
```{r}
i<-names(prds)[-1][17]
for(i in 2:lic){
tfi<-(as.vector(is.na(prds[,i,with=F])))
prds[(tfi),(i):=1]
} #prds[(tfi)] #prds[,(i),with=F]
prds$oneMeanVar<-apply(prds[,2:lic,with=F],1,mean,na.rm=T)
plot(prds$oneMeanVar,prds$meanVar)
plot(prds$oneMeanVar,prds$maxVar)
plot(prds$oneMeanVar,prds$countVar)
prds[(order(oneMeanVar))]$allColNm[1:30]
```
what was the varimp formula?
what the 0 -5 and huge + mean?
Bunch of nans and infs
Normalization is a pita; some runs had bigger RMSE ranges because of transform.
maybe should have used min change (mean , median , 0, permute) instead of permutation?
Goal is to approximate importance of this feature to overall model performance but seems like everything overshoots!
```{r}
aggsum<-data.frame(sum=sapply(prds[,2:lic,with=F], sum, na.rm=TRUE),
sd=sapply(prds[,2:lic,with=F], sd, na.rm=TRUE))
plot(aggsum$sd,aggsum$sum)
aggsum$sum[order(aggsum$sum)]
```
what percent of the usefullness of model caused by each feature?
kind of cramming broken data into seemingly good form
ok now I feel like standardizing; all less than 1 set to 1 , log
subtract 1 from all, divide all by column's sum
```{r}
i<-names(prds)[-1][17]
for(i in 2:lic){
tfi<-(as.vector((prds[,i,with=F])<1))
prds[(tfi),(i):=1]
intrm<-log(prds[,i,with=F])
intrm<-intrm/sum(intrm)
prds[,(i):=intrm]
}
prds$minVarN<-apply(prds[,2:lic,with=F],1,min,na.rm=T)
prds$maxVarN<-apply(prds[,2:lic,with=F],1,max,na.rm=T)
prds$meanVarN<-apply(prds[,2:lic,with=F],1,mean,na.rm=T)
```
now plot it ?
nope this data is useless. will come back to it whenever i fix and figure out whatever dalex did
```{r}
plot(prds$maxVarN,prds$meanVarN)
plot(prds$oneMeanVar,prds$maxVarN)
prds[(order(maxVarN))]$allColNm[1:30]
prds[(order(meanVarN))]$allColNm[1:30]
```
look for highly uncorrelated recipies
by task select single best model and add to possiblyUniq
then those models that most strongly discorrelate are also added
Since features should be uncorrelated? A rare column effectively used indicates uniqueness? or just cluster?
```{r}
```