forked from btgraham/SparseConvNet-archived
-
Notifications
You must be signed in to change notification settings - Fork 0
/
cifar10indexLearning.R
33 lines (30 loc) · 960 Bytes
/
cifar10indexLearning.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
a=as.matrix(read.table("CIFAR-10 train set.features"))
trainX=a[,-(1:2)]
#trainX=trainX*3*(trainX<0)+trainX*(trainX>0)
trainY=a[,2]
a=as.matrix(read.table("CIFAR-10 test set.features"))
testX=a[,-(1:2)]
#testX=testX*3*(testX<0)+testX*(testX>0)
testY=a[,2]
rm(a)
library(e1071)
s=svm(trainX[1:1000,],trainY[1:1000],type="C",kernel="lin",cost=1)
mean(predict(s,trainX)==trainY)
mean(predict(s,testX)==testY)
library(grid)
load("/home/ben/Archive/Datasets/cifar/cifar-10-R/cifar10.RData")
for (i in 1:dim(trainX)[2]) {
a=train.X[order(trainX[,i])[c(1:25,40000+9976:10000)],]/255
#a=test.X[order(testX[,i])[c(1:25,9976:10000)],]/255
a=array(aperm(array(a,c(5,10,32,32,3)),c(4,1,3,2,5)),c(32*5,32*10,3))
png(paste("index-learning/",i,".png",sep=""),height=32*5,width=32*10)
grid.raster(a,interpolate=F)
dev.off()
}
par(mfrow=c(2,5))
for (i in 0:9) {
a=cor(trainX,trainY==i)
b=cor(testX,testY==i)
plot(a[order(a)])
points(b[order(a)],col=2)
}